Category: Sizing Approaches and Logic

Posted on

Storage calculation for clusters with mixed capacity nodes

This article explains the logic for storage calculations for clusters having node with different storage capacities

What has changed?

Previously, the capacity calculations had been based on aggregate capacity across nodes in the cluster. This total capacity used to be the base for calculating usable and effective capacity in the cluster.

For example: Consider 3 nodes , N1=20TB, N2=20TB and N3=10TB

Based on the above, the total capacity available is 20+20+10 = 50TB and assuming (N+1) , the available nodes are N2+N3 = 30TB. Thus, 15TB can be used for data and 15TB for RF (assuming RF2)

With the new update: Sizer also ensures that the RF copy of the data and the data itself do not share the same node.

In the above example: after N+1, two nodes are available N2 = 20TB ,N3 = 10TB.

If we allow writing 15TB of data (and 15 for RF), part of the data and RF has to be on same node as N3 is only 10TB. So, to ensure the RF copy and data is on separate nodes, the usable storage in this case would be 20TB ( 10TB of data on N2 and its RF on N3 or vice versa).

Note: Although the same logic is used for both homogeneous and mixed capacity clusters, the difference is seen primarily for the mixed capacity clusters.

Here is a detailed write up on how the usable storage is calculated for clusters with mixed capacity nodes for different scenarios across RF2 and RF3

Algorithm for RF2

If we have only one node with non-zero capacity Cx, then in RF2 the replication is done between the different disks of the same node and hence the extent store in this case will be Cx / 2 (RF), else one of the below cases applies. Let us say, we have nodes with capacities C1, C2, C3, …., C[n] which are in sorted order according to their capacities. There are 2 cases to consider for RF2, to compute the effective raw storage capacity:

Case-1: C1 + C2 + C3 + …. + C[n-1] <= C[n] 
If this is the case, then the total amount of storage that can be replicated with a factor of 2 is  ∑(C1, C2, C3, …., C[n-1])

Case-2: C1 + C2 + C3 + …. + C[n-1] > C[n]
If this is the case, then the (total storage capacity) / 2 (RF) can be replicated among the available nodes. In other words, half the total capacity can be replicated.

Algorithm for RF3

Let us say, we have nodes with capacities C1, C2, C3, …., C[n] which are in sorted order according to their capacities. Algorithm for RF3 is slightly different from that of RF2 because we need to accommodate the replica of data on 2 nodes, as opposed to a single node on RF2.

  1. Since there are 3 replicas to place, we calculate the capacity difference between the 2nd largest (C[n-1]) and the 3rd largest (C[n-2]) entities as ‘diff’. This information is necessary so that given an optimal placement scenario where the first replica is placed on the entity with the smallest capacity, the second replica is placed on the entity with the largest capacity (C[n]) and the third replica is placed on the entity with the 2nd largest capacity (C[n-1]); the difference between the 2nd and the 3rd largest capacities ((C[n-1]) – (C[n-2])) will help us quickly deduce when the 2nd largest entity will become equal to the 3rd largest entity by virtue of space consumed on the former via replica placement.
  2. By deducting either the ‘diff’ calculated above (or) the capacity of the smallest entity and simulating RF3 placement such that C[n-2] and C[n-1] have now become equal (note that the difference between C[n] and C[n-1] will remain constant during this since the same capacity is deducted from both of them), in O(N) we arrive at the possibility of:
    • Case-1:  Only 3 entities remain with non-zero capacities, in which case the amount of data that can be accommodated among these 3 nodes with RF of 3 (one actual node and 2 replicas) is the smallest remaining capacity, which is C[n-2].
    • Case-2:There is capacity left in C[n-3] (i.e. the 4th largest entity) and any number of nodes before it (i.e., C[n-4], C[n-5], … etc) and C[n-2] == C[n-1] (i.e. the capacities remaining on the third and the second largest entities have become equal). This is because at this point, the capacity on the smallest entity remaining (the smallest non-zero entity before C[n-2] i.e) is greater than C[n-1] – C[n-2], indicating that after placing the first replica on C[n] and second replica on C[n-1], the time has come where the capacity on C[n-1] == C[n-2]. At this point, for the next bytes of data, the second replica will go to C[n] while the third replica will be round robin-ed between at least 2 (or more) entities. Now in this scenario as well, 2 cases can arise:
      • Case-2(a): (C1 + C2 + … + C[n-1]) / 2 <= C[n]
        Now, if C[n]’s capacity is so high that it means that for every 1st and 3rd replicas placed on the lowest capacities nodes upto C[n-1], the second replica always finds space on C[n], then it implies that, if (C1 + C2 + … + C[n-1]) / 2 <= C[n], then the amount of storage that can be accommodated on available nodes with RF of 3 is the lowest among the two sides of the above equation i.e., (C1 + C2 + … + C[n-1]) / 2, as we cannot consume the full space on C[n].
      • Case-2(b): (C1 + C2 + … + C[n-1]) / 2 > C[n]
        But if C[n]’s capacity is not so high as in case (a), i.e., (C1 + C2 + … + C[n-1]) / 2 > C[n], then replica placements for one of the replicas will be on the largest entity C[n], while the other two replicas will round-robin amongst the other largest capacity entities (since the capacities remaining on at least 2 entities C[n-2], C[n-1] are already equal). This will continue until C[n] becomes equal to C[n-1], which is guaranteed to happen eventually because the replicas consume space on C[n] at least at a rate double than C[n-1], C[n-2], … From that point, both the second and the third replicas will continue being round robin-ed across all the remaining entities, and thus all the capacities remaining at that point can be fully consumed. Hence, in this case, the amount of storage that can be accommodated is the sum of all remaining (non-zero) entities divided by 3 (RF).

Terminologies

We will explain the terminologies with an example. To keep it simple, let us take a homogenous cluster with 6 identical nodes.

Effective Raw Capacity is the Total Raw Storage available across clusters within the scenario. In this case, it is 6 nodes with 6 x 1.92TB drives, amounting to 69.12 TB or 62.86 TiB, as shown in the first row in Figure 2 below.

Failover Capacity Overhead is the storage that is discarded according to the failover plan selected. In the above case, the failover plan is set to “Standard (N+1),” and hence, we arrive at this value by discarding all the storage available within the node. The solution in the above example is an all-flash node with 6 x 1.92TB drives per node. The Failover Capacity Overhead amounts to 11.52TB per node or 10.48 TiB per node. The same can be seen in the second row in Figure 2 above.

Effective Raw Capacity After Failover = Effective Raw Capacity – Failover Capacity Overhead

In the above example, that is 62.86 TiB – 10.48 TiB = ~52.39 TiB

Effective Usable Capacity After Failover = ~95% of Effective Raw Capacity After Failover as AOS stops writing to the disk when the cluster utilization reaches 95%.

In the above example, that would be 52.39 * ~0.95 = 50.22 TiB

Extent Store = (Effective Usable Capacity After Failover – CVM) / Replication Factor

In the above example, with RF set to 2, it would be (50.22 TiB – 9.14 TiB)/2 = 20.54 TiB

Effective Capacity = Extent Store + Savings (Storage Efficiency & Erasure Coding)

In the above example, without saving, the Effective Capacity is 20.54 TiB.

If we set the Storage Efficiency to 18%, the Savings amount to 4.52 TiB, and hence the Effective Capacity would be 20.54 TiB + 4.52 TiB = 25.06 TiB. The same can be seen in the bottom part of Figure 2.

Please note that the numbers will vary for heterogeneous nodes when different capacity nodes exist within a cluster. The larger the variance of capacity across nodes, the Extent Store, and hence, the effective capacity will be reduced to ensure there is enough storage present across nodes to maintain copies of data as per the desired Replication Factor.

Posted on

SAP HANA sizing guidelines

Sizing for SAP HANA  

Introductory video of SAP HANA in the sizer (non-ST VPN connection required).

https://nutanixinc.sharepoint.com/:v:/t/solperf/solperf_library/ETb67HebgR9Kg32r7jxTO3wBdXoHuOCCjvK2oTasctOgIQ?e=z4Epr2

Note:

  • Only use for SAP HANA based applications, not legacy SAP applications which make use of MSSQL, Oracle etc.
  • Multi-node SAP scale-out is not supported (used for larger SAP BW instances)
  • This sizing process does not vary for scale-up between AHV & VMware implementations
  • No spinning disks are used within a Nutanix cluster being used for SAP HANA
  • Any questions, support, or areas not covered – please use the SAP Slack channel

Supported Platforms:

  • Only Dell, HPE, Fujitsu and Lenovo are supported for SAP HANA, not NX.
  • If another OEM is selected, SAP HANA will not be shown as an available workload

 

Defaults

  • RF2 is used (RF3 is under testing, so not selectable in Sizer)
  • Compression is disabled, and not typically of value for SAP HANA
  • Higher default CVM resource is reserved

HANA Inputs

  • NVMe can be added for higher IO loads, such as a high usage SAP S4/HANA
  • Cost/Performance largely drives cpu choice. Ideally an implementation’s potential compute load in SAPS would be known. Please reach out for support in estimating and reviewing such information.

Environments

There would typically be two environments within a Nutanix cluster where production and non-production are mixed. Production rules should be applied both to all production instances, and any other instances that should be treated as production. This might apply to a QAS/Test environment and will typically apply to any DR cluster.

Production:

  • For most SAP applications (e.g., production S4) there is an SAP HANA database, and one or more application server instances. Some uses of SAP HANA do not use an application server, in which case just use a small one in the sizing exercise.
  • In addition to the Application Server instances, and the SAP HANA database, a small VM called the ASCS is often called for. This ASCS would be around 2c/24GB RAM/100GB disk.
  • Generally, production has two or more application server instances. Typically, 2 – 6 cores, with around 24GB/core. Multiple instances for larger loads. Small storage space requirement for os & application image.
  • For a downtime requirement of less than 20 minutes, a pair of SAP HANA instances should be sized.
  • There is no over commit of cpu or memory
  • Servers must have all memory channels filled and balanced, so 6 or 12 DIMMs per cpu. – Sizer auto recommendation enforces this consideration
  • L suffix cpus are required for largest memory instances
  • Available storage for SAP HANA should be around 2.5x to 3x memory (3x is used in Sizer)
  • Production rules – SAP HANA instances are on whole dedicated cpus and so cannot be allocated to the CVM cpu
  • HANA System Relication(HSR) – is exactly a copy of the HANA VM. In Sizer, add another HANA VM if implementing the HSR.

 

Non Production:

 

 

  • QAS/Test landscape tends to match nonPRD for size of instance
  • If an operating system HA cluster is used in production, there is typically at least one such cluster outside of production also – used as a testbed.
  • Each SAP solution would normally have two or three non-production landscapes
  • Solution Manager (SolMan) is often overlooked, and not asked for. It is a required instance in the overall deployment and would be sized in PRD with one SAP HANA instance and an application server instance. Another such pair for QAS/test. No HA clustering would be required.
  • DEV, SBX etc. are usually subsets in memory size.

 

Feb 8, 2021

First release of SAP HANA in the sizer:

 

Posted on

Business Continuity and Disaster Recovery

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

BCDR

 1.  Has your organization done a Business Impact Analysis of your applications and workloads, and would you be able to share that with us?

Why ask? A business impact analysis should have rated the customer’s applications based on business criticality and may have informed them as to which are the most important applications and what level of data loss and time loss would be acceptable, what the impact in both revenue and non-revenue would be, what the systems and dependencies of those critical applications and processes are.  This information would be HIGHLY beneficial to have in designing a DR solution for your customer.

2.  If a Business Impact Analysis is not available, please breakdown applications and their corresponding recovery point objectives and associated recovery time objectives.  Also breakdown different retention requirements for on-prem and the recovery site.

Why ask? Recovery point objectives lets us know what Nutanix products can be used to meet the objective and the RTO (recovery Time Objective) may cause the design to shift depending on how fast the application needs to be back up and running. It is also good to get this backed by the business versus the IT department making the decisions. We should also mention that DR and Backup are different. The retention for DR should be as long as is needed to meet the RTO. If you have a very long RTO it is probably best served by a more cost effective backup solution.

 3.  For the applications listed please list the change rate for your applications. You can use existing backups and the deltas between them to find out the change rate. For any aggressive RPO that may require near-sync you will also want to know the write throughput for the application.

Why ask? Write throughput is the best indicator for network bandwidth throughput with near-sync.

 4.  What backup and/or DR products do you currently use?

Why ask? Helps to find any synergies or aid to move them to products which are more Nutanix friendly.  Example: if they are using Zerto, they may be great candidates for near-sync.

 5.  In the event of a failure, who is on-call to help with the restore. Will everyone be trained on recovery plans and procedures?

Why ask? This should help the customer to consider the operational implications of DR and help them see the need to simplify the recovery process. When bad events happen everyone needs to feel comfortable and empowered to help with the restoration.

 6.  What are you using today for Disaster Recovery and have you tested the solution to verify it works as expected?  How has the solution met or not met your requirements and expectations?

Why ask? Discovery to uncover the current topology and the how it is currently performing.

 7.  Are there any other systems that need to be taken into account for a proper disaster recovery plan that would be outside of the Nutanix infrastructure? (physical servers, mainframe, other physical devices)

Why ask? This is to make sure that the customer has thought through the entire scope necessary for a DR based on their requirements.

8.  What type of disasters are you planning for?

Why ask? Depending on the type of disaster scenario that could involve different products or operational procedures (for example, think of the difference of a disaster like a pandemic where all your systems are working, but your workforce can’t come in the office versus a geographic event like an explosion or earthquake in comparison to a ROBO office server closet being flooded, or even a ransomware attack).  Help the customer walk through different scenarios and possible reasons for needing to invoke a recovery plan and how we can help give them a recover in depth strategy (snapshots, backup, remote DC DR, stretch clusters, Cloud Provider DR, DRaaS with Xi Leap)

9.  Can you sites run independently? If so, how long?

Why ask? This helps to determine the criticality of each site

10.  What is your Retention Policy, and who made that decision (regulatory or self-imposed)?

Why ask? This helps to understand what is driving the need for retention and if it is internal or externally mandated.

11.  Do you have any regulatory requirements such as HIPAA, PCI, SEC, SOX?

Why ask? This will impact the particular design of the recovery solution, what features and controls will need to be in place (encryption, data sovereignty, RBAC, Logging, etc.)

12.  Do you need immutable copies of the data/VMs?

Why ask? These questions allow SE to determine sophistication of customers as well as how long they need the data. If there is a 7 year policy, customers will most likely need a 3rd Party backup tool in order to tier the data to an Object Store that supports WORM (Write Once Read Many).  This could also stem from requirements to help mitigate the risk of ransomware.

13.  Is this going to be a net new DR Solution?

Why ask? Do we have the flexibility to design a net new DR strategy?

14.  Will the source and destination clusters both be Nutanix clusters?

Why ask? If Nutanix is not the source and destination, then it will cause constraints for the design of the DR Solution.  (i.e. it will force hypervisor choice, replication can’t be array based but will need to be done with a separate software product like SRM or Zerto).  It may also require licensing changes

15.  Do you have a requirement for the backups/DR to be on separate hardware?

Why ask? Understand how to architect and select the infrastructure for the backup and DR targets

16.  What does your desired replication topology look like?

Why ask? Gives us the information on whether we need Professional or Ultimate AOS licenses and helps to map out the topology for replication (i.e. A->B->C; A->B; A->B and A->C).  This also allows us to discover if sites are active/passive and exactly what the definitions for active and passive mean (i.e. a data center with power, or VMs ready to power on, or VMs already powered on and able to switchover immediately, etc.)

17.  Does DR need to be the same performance as production?

Why ask? This helps to give an understanding of if DR is a checklist item for them or a significant business requirement (i.e. they’ll lose significant money in the event of any downtime of the production site).  Allows you to be able to size the solution appropriately.  Ensure to get sign off if they do decide to allow for DR to be undersized.

18.  Do you need to have replicated copies of all the VMs but only plan to restore a subset?

Why ask? Helps determine potential licensing requirements and size of target cluster.

19.  How often do you test your disaster recovery plan and what does the plan look like?

Why ask? This will help with understanding if the customer has actually validated any disaster recovery plan that they have implemented to ensure that it will actually work.  Oftentimes folks are optimistic with regards to how well their plan will actually work, so having tested it brings a sense of reality to the plan and can help them course correct.

20.  Do you have existing runbooks that are used in the event of a disaster?

Why ask? Knowing this will help us understand their processes in the event of a disaster.  Also, if they don’t have any runbooks this will let us know we may need to help them out in putting a plan together that they can use and test.

21.  Databases: How are you currently backing up and protecting your database environments?

Why ask? This can give us information about how they are protecting their most critical assets.  They may already be using or licensed for database level replication or clustering technology which would give them more granular levels of control than a traditional storage replicated VM.  This can also help as part of the discussions around RPO and RTO for these more critical systems which may need higher levels of availability than other parts of the infrastructure. 

22.  What hypervisor(s) are you using? Are you open to cross-hypervisor DR?

Why ask? This will let us know what replication products we can use (if the source or destination cluster is non-Nutanix) and can open up the possibility of leveraging AHV for the target cluster if they are using ESXi on Nutanix as the source/primary cluster. 

23.  Do you have a separate network for DR traffic? Do you require encryption on those links?

Why ask? This helps with understanding if network segmentation is necessary to be configured in Prism for DR Replication traffic and whether or not the customer needs to supply encryption in flight for that network.  

24.  What is the current bandwidth between sites that you plan to use for DR replication? Also, what is the latency between those sites?

Why ask? We need to know how big the pipes are between sites so that we can ensure that the RPO the customer has defined as their requirement will be able to be met based on the rate of change.  Also ensuring that the latency between sites meets the minimum requirements listed for Metro Availability or the Metro Witness.  

25.  What is the current rate of utilization of the network links between the sites you plan to use for replication traffic?

Why ask? These links may be used for other traffic and could impact the available bandwidth that you assume you will have access to for replication traffic.  See if you can get utilization over a 30 day period, and if possible over several months to see any trends of increase or decrease in utilization.

26.  How do you handle IP addresses on the recovery site for VMs that have failed over?

Why ask? This allows us to discover what type of networking failover scenario(s) the customer would prefer to use: Overlay Networks; Stretched Layer 2 subnet between sites; perform a full subnet failover from the primary to secondary site by updating routes to point to the new recovery site; allow IP addresses to change when failed over (this can cause obvious challenges of broken applications that hard code IP addresses, updating of DNS and cache entries, etc.).

Posted on

Files

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

Files

1.  Is this replacing a current solution, or is this a net new project?
     a.  What’s the current solution?

Why ask? This question helps us understand the use case, any current expectations and what the competitive landscape may look like.

2.  Using an existing Nutanix cluster (with existing workload) or net new Nutanix cluster?

Why ask?  If we’re sizing into an existing cluster we need to understand current hardware and current workload.  For licensing purposes adding Files to an existing cluster means the Files for AOS license. A common scenario has been to add storage only nodes to an existing cluster to support the new Files capacity.  If sizing into a new cluster we can potentially dedicate this cluster to Files and use Files Dedicated licensing.

3.  Is this for NFS, SMB or both?
     a.  Which protocol versions (SMB 3.0, NFSv4, etc)?

Why ask?  We need to understand protocol to first validate they are using supported clients.  Supported clients are documented in the release notes of each version of Files.  Concurrent SMB connections also impact sizing with respect to the compute resources we need for the FSVMs to handle those clients.  Max concurrent connections are also documented in the release notes of each version.  It also helps us validate supported authentication methods.  For SMB, we require Active Directory where we support 2008 domain functional level or higher (there is no local user or group support for Files).  For NFS v4 we support AD with Kerberos, LDAP and Unmanaged (no auth) shares.  For NFS v3 we support LDAP and Unmanaged.

 4.  Is there any explicit performance requirement for the customer?  Do they require IOPs or throughput or latency numbers?

Why ask?  Every FSVM has an expected performance envelope.  There is a sizing guide and performance tech note on the Nutanix Portal which give a relative expectation on the max read and write throughput per FSVM and max read or write IOPs per FSVM.  Throughput based on reads and writes are integrated into Nutanix Sizer and will impact the recommended number of FSVMs. https://portal.nutanix.com/page/documents/solutions/details?targetId=TN-2117-Nutanix-Files-Performance:TN-2117-Nutanix-Files-Performance

5.  Do they have any current performance collection from their existing environment?
      a.  Windows File Server = Perfmon
      b.  Netapp = perfstat
      c.  Dell DPACK, Live Optics

Why ask?  Seeing data from an existing solution can help validate the performance numbers so that we size accurately for performance. 

6.  What are the specific applications using the shares?
       a.  VDI (Home Shares)
       b.  PACS (Imaging)
       c.  Video (Streaming)
       d.  Backup (Streaming)

Why ask?  When sizing for storage space utilization the application performing the writes could impact storage efficiency.  Backup, Video and Image data are most commonly compressed by the application.  For those applications we should not include compression savings when sizing, only Erasure Coding.  For general purpose shares with various document types assume some level of compression savings.  

7.  Are they happy with performance or looking to improve performance?

Why ask?  If the customer has existing performance data, it’s good to understand if they are expecting equivalent or better performance from Files.  This could impact sizing, including going from a hybrid to an all flash cluster. 

 8.  How many expected concurrent user connections?

Why ask? Concurrent SMB connections are a required sizing parameter.  Each FSVM needs enough memory assigned to support a given number of users.  A Standard share is owned by one FSVM.  A distributed share is owned by all FSVMs and is load balanced based on top level directories.  We need to ensure any one FSVM can support all concurrent clients to the standard share or top level directory with the highest expected connections. We should also be ensuring that the sizing for concurrent connections is taking into account N-1 redundancy for node maintenance/failure/etc.

9.  Will the underlying hardware config support larger or more FSVMs if additional throughput or performance is required?

Why ask? Files is a scale-out and scale-up workload so you need to know what growth in the environment can look like.

 10.  Current share configuration including number of shares?

Why ask?  Files has a soft (recommended) limit of 100 shares per FSVM.

11.  Directory structure:
       a.  Large number of folders in share root?

Why ask?  This indicates a large number of top level directories making a distributed share a good choice for load balancing and data distribution.

       b.  Files in share root?

Why ask?  Distributed shares cannot store files in the share root.  If an application must store files in the root then you should plan for sizing using standard shares.  Alternatively, a nested share can be used. 

       c.  Total size of largest single directories?

Why ask?  Nutanix supports standard shares up to 140TB.  And top level directories in a distributed share up to 140TB.  These limits are based on the volume group supporting the standard share or top level directory.  We need to ensure no single folder or share (if using a standard share) surpasses 140TB. Files Compression can yield more usable storage per share as well. Nutanix Files – Deployment and Upgrade FAQ https://portal.nutanix.com/page/documents/kbs/details?targetId=kA00e000000LMXpCAO

       d.  Largest number of files/folders in a single folder?

Why ask?  Nutanix Files is designed to store millions of files within a single share and billions of files across a multi-node cluster with multiple shares.  To achieve speedy response time for high file and directory count environments it’s necessary to give some thought to directory design. Placing millions of files or directories into a single directory is going to be very slow in file enumeration that must occur before file access.  The optimal approach is to branch out from the root share with leaf directories up to a width (directory or file count in a single directory) no greater than 100,000.  Subdirectories should have similar directory width.  If file or directory counts get very wide within a single directory, this can cause slow data response time to client and application.  Increasing FSVM memory up to 96 GB to cache metadata can help improve performance for these environments especially if designs for directory and files listed above are followed.

12.  Total storage and compute requirements including future growth?

Why ask?  Core sizing question to ensure adequate storage space is available with the initial purchase and over the expected timeframe. 

13.  Does your sizing include the resources to run File Analytics?

Why ask?  FA is a key differentiator for Files, and drives a lot of customer delight and insights into their data. Every SE and sizing should assume that any Files customer will want to run FA as well – don’t present as an optional component.

14.  Percent of data considered to be active/hot?

 Why ask?  Understanding the expected active dataset can help with sizing the SSD tier for a hybrid solution.  Performance and statistical collection from an existing environment may help with this determination.

 15.  Storage change rate?

Why ask?  Change rate influences snapshot overheads based on retention schedules.  Nutanix Sizer will ask what the change rate is for the dataset to help with determining the storage space impact of snapshot retention.

 16.  Any storage efficiency details from the current environment (dedup, compression, etc.)?

Why ask?  Helps to determine if data reduction techniques like dedup and compression are effective against the customers data.  Files does not support the use of deduplication today, so any dedup savings should not be taken into account when sizing for Files.  If the data is compressible in the existing environment it should also be compressible with Nutanix compression.

 17.  Block size of current solution (if known)?

Why ask?  Block size can impact storage efficiency.  A solution which has many small files with a fixed block size may show different space consumption when migrated to Files, which uses variable block lengths based on file size.  For files over 64KB in size, Files uses a 64KB block size.  In some cases a large number of large files have been slightly less efficient when moved to Nutanix Files.  Understanding this up front can help explain differences following migrations.

18.  Self Service Restore (SSR) requirements (share level snapshots)?

Why ask?  Nutanix Files uses two levels of snapshots, SSR snapshots occur at the file share level via ZFS.  These snapshots have their own schedule and Sizer asks for their frequency and change rate under “Nutanix Files Snapshots.”  The schedule associated with SSR and retention periods will impact overall storage consumption. Nutanix Files Snapshots increase both the amount of licensing required and total storage required, so it’s important to get it right during the sizing process.

 19.  Data Protection/Disaster Recovery requirements (File Server Instance snapshots):
         a.  Expected snapshot frequency and retention schedule (hourly, daily, weekly, etc.)?

Why ask? Data Protection snapshots occur at the AOS (protection domain) level via the NDSF.  The schedule and retention policy are managed against the protection domain for the file server instance and will impact overall storage consumption.  Sizer asks for the local and remote snapshot retention under “Data Protection.”
Files supports 1hr RPO today and will support near-sync in the AOS 5.11.1 release in conjunction with Files 3.6.  Keep in mind node density (raw storage) when determining RPO.  Both 1hr and near-sync RPO require hybrid nodes with 40TB or less raw or all flash nodes with 48TB or less raw.  Denser configurations can only support 6hr RPO.  These requirements will likely change so double check the latest guidance when sizing dense storage nodes. Confirm that underlying nodes and configs support NearSync per latest AOS requirements if NearSync will be used.

        b.  Active/Active requirements (Peer Software)?

Why ask?  If the customer needs active/active file shares in different sites which represent the same data, we need to position a third party called Peer Software.  Peer performs near real time replication of data between heterogenous file servers.  Peer utilizes Windows VMs which consume some CPU and memory you may want to size into the Nutanix clusters intended for Files.

 20.  Feature Requirements:
         a.  Auditing? Which vendors?

Why ask?  Nutanix is working to integrate with three main third-party auditing vendors today, Netwrix (supported and integrated with Files), Varonis (working on integration) and Stealthbits (not yet integrated).  Nutanix Files also has a native auditing solution in File Analytics.
Along with ensuring audit vendor support, a given solution may require a certain amount of CPU, Memory and Storage (to hold auditing events).  Ensure to include any vendor specific sizing in the configuration.  File Analytics for example could require 8vcpu 48GB of memory and 3TB of storage.

         b.  Antivirus? Which vendors?

Why ask? Files supports five main Antivirus vendors today with respect to ICAP integration, McAfee, Symantec, Kaspersky, Sophos and Bitdefender.  If centralized virus scan servers are to be used you will want to include their compute requirements into sizing the overall solution.

         c.  Backup? Which vendors?

Why ask?  Files has full change file tracking (CFT) support with HYCU and Commvault.  Veritas, Rubrik and Veeam are or will soon be working on integration.  Other vendors can also be supported outside of CFT support.  If including a backup vendor on the same platform, you may need to size for any virtual appliance which may also run on Nutanix.

         d.  Multiprotocol? SMB + NFS? * (Engage with a Files Specialist/Solutions Architect if this is a customer requirement)

Why ask?  Multiprotocol is challenging, and often behaves differently than a customer imagines it will. One protocol is defined as authoritative and the other protocol maps onto it. If the customer does not already use multiprotocol shares and have a strong command of the technology, engage your SA to assist on the design to ensure success.

21.  Using DFS (Distributed File Server) Namespace (DFS-N)?

Why ask?  Less about sizing and more about implementation.  Prior to Files 3.5.1 Files could only support distributed shares with DFS-N.  Starting with 3.5.1 both distributed and standard shares are fully supported as folder targets with DFS-N.

 22.  Tiering requirements?

Why ask?  Files is targeting support for tiering in the 1H of CY21.  Tiering in this context means automatically moving data off Nutanix Files and to an S3 compliant object service either on-premises or in the cloud.  In scoping future requirements, customers may size for a given amount of on-premises storage and a larger amount of tiered storage for longer term archive.

23.  Access-Based Enumeration (ABE) Requirements?

Why ask?  Nutanix Files supports Access-based Enumeration (ABE). Is it a requirement to hide objects (files and folders) from users who don’t have NTFS permissions (Read or List) on a network shared folder in order to access them?  If so, we fully support it. 

24.  Reality Check: Files Mixed vs Dedicated clusters

Why ask?  Always double check the cost of Dedicated vs Mixed clusters. Dedicated can often be more cost effective, and accommodates larger FSVM sizes since the FSVMs  are capable of using the full amount of compute resources available to the cluster.

25.  Reality Check: Dedicated Cluster hardware minimums

Why ask?  Remember that Files is still a virtualized workload, so don’t assume the minimum possible hardware spec. Use 12 core 4214 CPUs as a reasonable minimum, or 14 cores if NearSync requirements dictate. 128GB memory per node will not allow for AHV + CVM + maximum FSVM size deployments, so consider 192GB, or nodes that can expand to 192GB after deployment.

26.  Reality Check: Implementation

Why ask?  Have a high level design of how you’ve designed/sized Files in your solution and communicate the design to the installer. Poor implementation, and implementation that doesn’t match the planned design, is one of the leading causes of customer satisfaction issues for Files.

27.  Reality Check: Files Prerequisites

Why ask?  Ensure that you’ve reviewed the relevant prerequisites and shared with the customer before deploying (Active Directory if using SMB, AHV/ESXi only – no Hyper-V, have a second VLAN if customer wants iSCSI isolation, Backup clients like Rubrik deployed in wrong subnet if using two networks)

28.  Reality Check: Clients

Why ask?  Review the list of supported Files clients and share with the customer. Laptops and desktops are rarely a problem, but document senders/multifunction printers that are used to scan paper and convert to PDFs on a file share can often be capped at only SMBv1 support, which Files does not and will never support.

Resources:

Xpert Storage Team Page:  https://sites.google.com/nutanix.com/americas-xpert-program/storage?authuser=1

Files Sales Enablement Page: https://sites.google.com/nutanix.com/files/home?authuser=1

Calls to action/next steps:

For a peer review of a sizing or to request meeting support after the Files first call is completed: create a SFDC opportunity and request a Storage solutions architect on the opportunity

Test Drive – Storage: https://www.nutanix.com/one-platform?type=tddata

Files Bootcamps: https://confluence.eng.nutanix.com:8443/display/SEW/Bootcamps (Internal Only)

Posted on

Server Virtualization

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

Server Virtualization

Generic & State of The Union Questions

1.  What does your server virtualization environment look like today?

2.  What do you find most challenging about your current environment?

3.  What about your environment keeps you up at night?

4.  What is working well in your current environment that you want to ensure is continued?

5.  What does your cloud strategy currently look like?
       a.  Cloud First, Multi-Cloud, Hybrid-Cloud, Cloud Stinks
             i.   And why has this strategy been chosen?
             ii.  Who is directing/championing this cloud strategy?

6.  Where do you want to be related to the cloud over the next 1-3 years?
       a.  Is there a preference of Cloud Provider? I.e. AWS, Azure or GCP?

Architecture/Solution Specific Questions

1.  What is your current x86 server vendor standard?
      a.  Are you happy with this vendor?
      b.  If so, what do you enjoy/appreciate most?
      c.  If not what do you find most challenging?

2.  What is your preferred storage vendor for Virtualization?
      a.  Are you leveraging RDM (Raw Device Mappings) for your workloads?
      b.  If so, can you please provide some workload examples?
            i.  Oracle, MS CSV’s, SCSI-3 Shared Devices

3.  What is your preferred storage vendor for Physical Workloads?

4.  How do you currently connect your storage to your servers?
      a.  NFS, FC, FCOE, iSCSI

5.  What SAN/Storage hardware is in place today?
a.  HDD/Hybrid/All Flass/etc.?
b.  How many spindles of each?
c.  How many Controllers/Storage Processors?

6.  What does the logical disk layout look like?
a.  RAID Level?
b.  Number of Disks per RAID Group?

7.  Who is your preferred Hypervisor vendor and what version are you using?

8.  How open would you be to considering other hypervisors?

9.  What is your preferred Networking Vendor?

10.  Are you leveraging Traditional 3-Tier Networking or Leaf-Spine Networking?

11.  What does your networking architecture/rack design look like?

12.  Are you integrating hypervisor networking and what are your current networking standards?
         a.  ACI/NSX/Arista/Cumulus

13.  A collection of data from your current environment is needed.  Can we use Nutanix Collector, RVTools (VMware), Dell LiveOptics, Microsoft MAP, Oracle AWR, or any other inventory collection you may have used to gather at a minimum, the following information?:
       a.  # of Virtual Machines
       b.  # of vCPUs
       c.  Current vCPU to Physical Core oversubscription
       d.  Current Physical CPU Model In hosts (for SpecInt Sizing/Comparison- http://ewams.net/?view=How_to_Size_the_CPUs_of_New_Systems_Using_my_Specint_Rated_Tool )
       e.  Allocated memory
       f.  Provisioned storage
       g.  Consumed storage
       h.  Largest vCPU allocation (for NUMA design)
       i.  Largest Memory allocation (for NUMA design)
       j.  Working set size (what will sit in SSD for Hybrid)-can be determined from daily incremental backups (https://www.joshodgers.com/2014/09/25/rule-of-thumb-sizing-for-storage-performance-in-the-new-world/ )

Backup & Data Protection Questions

1.  What is your primary method for backing up Virtual Machines?
      a.  RPO/RTO

2.  What is your primary method for backing up Physical Servers?
      a.  RPO/RTO

3.  What is your primary method for backing up In-Guest or In-OS applications?
      a.  Veeam/Commvault/HYCU/Networker/HP Data Protector

4.  Who is your preferred backup storage vendor
       a.  DataDomain/Netapp/Commvault-Hyperscale/Cohesity/Rubrik/DAS-JBOD/etc.

5.  What is your backup storage strategy
       a.  Retention On-Prem, Drain to Cloud, ETC
       b.  Do you utilize the GFS, or Grandfather-Father-Son retention policy for your backup rotation scheme and if so, how many of each are kept? For example, 10 daily, 5 weekly and 1 monthly.

6.  What is the average daily data-change rate (est. 30 days) identified through incremental backups?  ( https://www.joshodgers.com/2014/09/25/rule-of-thumb-sizing-for-storage-performance-in-the-new-world/ )

7.  Can you walk me through your recovery process?
      a.  Let’s assume you need to restore a VM, what are the steps, how long would it take?
      b.  Let’s assume you needed to restore a set of files or a database table, what are the steps and how long would it take?
      c.  Let’s assume you lost all your shared storage, how would you recover and how long would it take?

Disaster Recovery / Business Continuity Questions

1.  Do you have a Disaster Recovery / Business Continuity Plan?
     a.  If yes, can you Share? (If they do have a plan, but cannot give you a copy of the plan, then ask if they can step you through the plan they have defined.
     b.  If no, would you consider working with Nutanix to help develop this with you?

2.  For Virtual Machines are you leveraging any specific SW solutions to help provide DR for your environment?

3.  For your Physical Servers

4.  Walk me through the process of recovering your operations at a DR Datacenter.

5.  Are there any Active/Active Workloads?
      a.  If so, can they define what Active/Active exactly means to them.  Is Active/Active a running secondary datacenter with tape backups available for restore, or data replicated to another SAN but no automated/orchestrated recovery, or is it a Metro/Stretch Cluster across sites, etc.?

 

 

Posted on

Calm

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

Calm

Discovery Questions

1.  How are you currently automating IT Service Delivery today, do you have any:
     a.  IAAS – Infrastructure as a service
     b.  PaaS – Platform as a service
     c.  SaaS – Software as a service

Why ask?  It helps us understand the customer’s maturity level when it comes to application deployment and could uncover some of the competitive infrastructure.  See some of the possible competitive or other products we may be able to work with or integrate with.

2.  Are standardizing and compliance important to you in your IT Automation Delivery Strategy?
a.  Do you currently use a business intake or self-service request process via a solution such as ServiceNow, Cherwell, Remedy, etc. to automate IT service delivery?

Why ask? Gives us the opportunity to discuss our SNOW plugin. Also helps understand which front end they will use for the Calm implementation.

3. Do you have any contracts with the cloud providers (AWS, Azure or GCP)?
a.  What are the specific use cases or workload profiles consumed from the cloud providers?

Why ask? Helps us understand which providers they may consume with Calm. Helps us understand which services are still on-prem and available as a target for AOS. May help position Beam. Also helps us understand if they have a Microsoft EA which may force their spend to go to Azure.

4.  Can you describe the process, do you have any documentation  for VM, OS or Database Deployment and Management?

Why ask? It helps uncover their current pain points and possibly competitive landscape.  (This would typically be asked when talking to the Infrastructure Team) If the process is already well documented/defined, the hardest part of the implementation is already done.

5.  What tools do you leverage to automate your Windows or Linux Server builds beyond the imaging / template / cloning process?
      a.  vRA
      b.  Terraform
      c.  Puppet
      d.  Chef
      e.  Ansible
      f.  Salt
      g.  SCCM

Why ask? Helps understand the competitive landscape as well as integration points that will need to be solved

6.  How many VMs are under management today?

Why ask? It helps us estimate the size of the deal for licensing

7.  What does your infrastructure footprint for managing/running containers look like?
      a.  What tools are you using?
      b.  How many containers?
      c.  If Kubernetes, how many pods and containers?
      d.  Which version of Kubernetes? (AKS/EKS/Anthos/Openshift/Tanzu/etc)

Why ask? Helps understand their current place on the journey to cloud native apps.  If they are still investigating, we have an option to position Karbon. If they are using another product already, we may be able to provide the infrastructure for that environment. 

8.  Do you have any contracts with the cloud providers (AWS, Azure or GCP)?
      a.  What are the specific use cases or workload profiles consumed from the cloud providers?

Why ask? Helps us understand which providers they may consume with Calm. Helps us understand which services are still on-prem and available as a target for AOS.  May help position Beam.  Also helps us understand if they have a Microsoft EA which may force their spend to go to Azure.

9.  Are standardizing and compliance important to you in your IT Automation Delivery Strategy?
      a.  Do you currently use a business intake or self-service request process via a solution such as ServiceNow, Cherwell, Remedy, etc. to automate IT service delivery?

Why ask? Gives us the opportunity to discuss our SNOW plugin. Also helps understand which front end they will use for the Calm implementation. 

10.  In your application development organization, is Continuous Integration/Continuous Delivery (CI/CD) an operating principle?
      a.  What tools do you leverage in your current/targeted pipeline?
            i.    Jenkins
            ii.   Atlassian Bamboo
            iii.  CircleCI
            iv.  GitLab CI/CD
            v.   Azure DevOps

Why ask? Helps us understand the integrations needed for a successful implementation.

Resources:

Glossary of Terms: https://github.com/nutanixworkshops/calmbootcamp/blob/master/appendix/glossary.rst 

xPert Automation team page: http://ntnx.tips/xPertAutomation (Internal Only)

LinkedIn Learning – DevOps Foundations Learning Plan: https://www.nutanixuniversity.com//lms/index.php?r=coursepath/deeplink&id_path=79&hash=2ce3cb1f946cc3770bd466853e68ee36ddbcf5e1&generated_by=19794

Udacity+Nutanix: Hybrid Cloud Engineer Nanodegree

Calls to action/next steps:

1.  Create a SFDC opportunity, quote a Calm+Services bundle, add a DevOps resource request
2.  Test Drive: Automation
3.  Calm bootcamps (+Karbon, +CI/CD, etc.) (Internal Only)

Posted on

End User Computing

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

EUC – End User Computing

Basic discovery

 1.  What is the expected type of EUC workload?

Why Ask? Are we talking about VDI (Full Desktop), RDSH (Shared Desktop), Application Virtualisation (like Citrix Apps or Horizon ThinApp).  Please keep in mind that you need to ask this question for every different workload the customer needs. In most EUC projects there is not just one type of user requirement. You will find a lot of mixed workloads, like persistent and non-persistent desktops as well as application virtualization.
VDI-Core licensing can help if the customer wants to run resource intensive VDI workloads, like developers or VDI with vGPUs and DR scenarios. For more information on VDI licensing check: https://docs.google.com/document/d/1VkFYvpOsAqmbwCwPWfGxg9bFhAmsPyANocn9rwG6eQQ/edit

 2.  Which vendor is used to manage and broker Desktops and Apps?

Why ask? The main vendors are Citrix, VMware, Nutanix. Every vendor has its own transport protocol, which makes a difference in CPU usage. Citrix = ICA, VMware = Blast or PCoIP, Frame = Frame Remoting Protocol.

3.  What is the expected Operating System Version?

Why ask? Every new version of Windows has higher requirements on CPU and Memory. It can make a big difference, if an older Windows version compared to the latest version. Windows 10 Performance Impact Analysis: https://portal.nutanix.com/page/documents/solutions/details?targetId=TN-2113-Windows-10-Performance-Impact:TN-2113-Windows-10-Performance-Impact

4.  What type of Windows license model is in use (KMS versus MAK)?

Why ask?Depending on the solution that is determined you may need to have one license model or the other (i.e. Frame requires KMS)

5.  What Office version is used?

Why ask? For Microsoft Office it is the same issue, like Windows versions. Newer versions need more resources.  Office 2019 Performance Impact Results from LoginVSI: https://www.loginvsi.com/login-vsi-blog/98-login-vsi/907-office-2019-performance-impact

6.  What other applications will be used?

Why ask? The applications used are having have a strong impact on the CPU. There might be single-threaded applications, which need high clock speed. Or you run applications that are multi-threaded and need a higher core count.

7.  What is the expected type of user?

Why ask? In sizer we ask for user types: task, knowledge, power user or developer. Every user type comes with a specific workload profile – memory, # vCPUs, vCPU:pCPU ratio, disk size. Sometimes the customer can give you details on the VM sizing, but not on the expected vCPU:pCPU ratio. Depending on the workload expected you can set the ratio. For more information read here: https://www.loginvsi.com/documentation/index.php?title=Login_VSI_Workloads

8.  How many concurrent users will you have on that workload?

Why ask? Concurrent user defines the number of active users. How many VMs need to run at the same time. Our VDI licensing is based on active VMs. You can have more users in an environment, but they can share resources, if they don’t work on the platform at the same time.  This can impact how you size compute and memory, but remember that storage may be needed for all possible users. 

9.  What provisioning method is used?

Why ask? Depending on the workload the VMs can be persistent or non-persistent. Persistent desktops will be treated like normal VMs. Non-persistent VMs will have a different storage footprint, since they are sharing a single boot disk and have additional write cache disks, which will be deleted after a VM reboots. Citrix uses MCS (Machine Creation Service) or PVS (Provisioning Service). VMware uses LinkedClones or InstantClones.

10.  Where are the user profiles stored?

Why ask? Using our own Files solution we can provide storage for the user profiles. Today you will mostly meet FSLogix profile containers, which still need an SMB share to be stored on and loaded during user logon.

11.  Do you need additional GPU support?  * (This may warrant engaging with a Solutions Architect or EUC Specialist for proper sizing and configurations)

Why ask? In order to accommodate applications like CAD or requirements in number of monitors and high resolution you need to add NVIDIA GPUs. An overview of vGPU Profiles can be found here: Virtual GPU Software User Guide :: NVIDIA Virtual GPU Software Documentation https://portal.nutanix.com/page/documents/solutions/details?targetId=SN-2046_vGPU_on_Nutanix:SN-2046_vGPU_on_Nutanix

 12.  Are there any other special requirements?

Why ask? Does the customer need RF3? Block Awareness, Rack Awareness, Storage encryption or replication)

VM Details

 1.  How many vCPUs?

Why ask? The number of vCPUs impacts the performance of the VM and the density of the host. Solution Engineering found the sweet spot to be at 3 vCPUs for a Windows 10 desktop or 8 vCPUs for Windows Server based Desktops

2.  What is the ratio between vCPU to pCPU?

Why ask? See question 6

3.  What is the requested CPU size per User?

Why ask? How much MHz does every user need to run the workload? It is very rare a customer can answer this. It is more common in application virtualization environments, since a number of users share the same VM and its resources.

4.  What CPU is currently in use?

Why ask? If the CPU currently used can handle, you could choose the same clock speed. But keep in mind in a virtualized environment resources are shared and you might have additional tasks running, like Files, which can impact the CPU.

5.  How much Memory is required per VM?

6.  What is the disk size?

Why ask? Depending on the provisioning method used you need the size of the Master Image (also called parental disk or Sandbox) and the Write Cache per VM. The write cache stores the temporary files written during the VM is active. For persistent VMs you need the disk size of the Master Image, which is then cloned into separate VMs.

7.  Are you planning to use microsegmentation to secure your VMs? If yes, what solution will be used?  * (This may warrant engaging with a Flow or Networking specialist or Solutions Architect)

Why ask? Position Flow on AHV or remember to size for a NSX appliance on every host.

8.  Are you planning on using an App layering solution?

Why ask? This saves the customer the need to manage lots of different master images. With App layering you have one master image and when a user logs the system will automatically attach additional disks, which contain the required applications. We can use shadow cloning to make those disks available locally.

General supporting Infrastructure

 1.  What Hypervisor are you planning to use?

Why ask? Different Hypervisors have different needs. If the customer chooses AHV we need to plan for PC, if it is VMware we may need to accommodate vCenter and for HyperV we need SCCM.

2.  Where will generic required services run?

Why ask? By generic services we mean AD, DHCP, DNS, Printing, licensing or application backend services. Some of them might be running in the cloud or on existing infrastructure. If running on the Nutanix Cluster take note of size and see Server virtualization questions.

 3.  Where will user profiles, home shares or App disks (if used) be stored?

Why ask? This is an opportunity to position Files. Usually customers today use a profile container to store user profiles. FSLogix is the most common solution used by customers, since it is included in their licensing. Please be aware that Files Services running on the same cluster do have a performance penalty during logon times.

4.  What is your DR strategy?

Why ask? Every customer needs a DR strategy for their EUC environment. A great question to position Xi Nutanix Clusters on AWS, replication and our unique VDI licensing approach.
Also need to calculate additional resources in your sizing, depending on the customers strategy.

Citrix Infrastructure

 1.  Where do you plan on running your Citrix services?

Why ask? Customers can choose to run all Citrix related services (like Studio, Databases, Storefront) as service in the cloud, managed by Citrix or on premises. If the customer chooses to run the services on premise, he can still run it on a different infrastructure. If he chooses to run it on the same cluster, please size additional server virtualization VMs. Guidelines on VM requirements can be found here: https://docs.citrix.com/en-us/citrix-virtual-apps-desktops/system-requirements.html
If the customer chooses Citrix Virtual Apps and Desktop (CVAD) Service you still need an additional Windows server as Cloud Connector.

A typical on premise implementation would need the following servers:
SQL deployment. What type of HA? (Always On, SQL Clustering with WSFC)
StoreFront Servers (HA, N+1)
Citrix Desktop Studio and Director (N+1)
Optional Provisioning Server (PVS) (N+1)
Network Load Balancer
Global Server Load Balancing
Profile Management Infrastructure (File services)
AppLayering Infrastructure
Any Endpoint Management technologies?
Failure domain sizes (Prism Central sizing)
Dedicated Infra Management Cluster or part of the Citrix cluster

VMware Infrastructure

1.  Where do you plan on running your VMware Horizon View Services?

Why ask? Customers can choose to run all Horizon related services as service in the cloud, managed by VMware or on premises. If the customer chooses to run the services on premise, he can still run it on a different infrastructure. If he chooses to run it on the same cluster, please size additional server virtualization VMs. Guidelines on VM requirements can be found here: https://docs.vmware.com/en/VMware-Horizon-7/7.12/horizon-installation/GUID-858D1E0E-C566-4813-9D53-975AF4432195.html

A typical on premise implementation would need the following servers:
SQL deployment. What type of HA? (Always On, SQL Clustering with WSFC)
Unified Access Gateway Appliances (N+1)
vCenter (N+1)
Horizon Connection Server (N+1)
Optional View Composer (N+1)
Profile Management Infrastructure (File services)
AppLayering Infrastructure
Any Endpoint Management technologies?
Failure domain sizes (Prism Central sizing)

Xi Frame Infrastructure

1.  Will there be Xi Frame User VMs on premise?

Why ask? For the use of Xi Frame on premise you need to add additional server VMs.
Cloud Connector Appliance (CCA)
Secure Gateway Appliance (SGA)

2.  Do you currently use Prism Central today and does it run in the same Prism Element cluster where the Frame workloads will run?

Why ask? Prism Central is required for Frame workloads for on premises Frame deployments

Advanced Discovery

1.  How do you optimize your image?

Why ask? Image optimization is crucial in all EUC environments. The host density and the user experience is increased by optimizing the VM using the tools provided by Citrix, VMware or vendor independent.
Citrix: https://support.citrix.com/article/CTX224676
VMware: https://flings.vmware.com/vmware-os-optimization-tool

2.  What is your Antivirus Strategy?

Why ask? The right AV solution can also have a massive impact on user experience and host density. If not done correctly all file operations lead to file scans, which increase CPU and IO on the host.

 

Posted on

Databases

Introduction – Please Read First

These questions are here to assist with ensuring that you’re gathering necessary information from a customer/prospect in order to put together an appropriate solution to meet their requirements in addition to capturing specific metrics from tools like Collector or RVTools. 

This list is not exhaustive, but should be used as a guide to make sure you’ve done proper and thorough discovery.  Also, it is imperative that you don’t just ask a question without understanding the reason why it is being asked.  We’ve structured these questions with not only the question that should be asked, but why we are asking the customer to provide an answer to that question and why it matters to provide an optimal solution. 

Questions marked with an asterisk (*) will likely require reaching out to a specialist/Solution Architect resource at Nutanix to go deeper with the customer on that topic/question.  Make sure you use the answers to these questions in the Scenario Objectives in Sizer when you create a new Scenario.  These questions should help guide you as to what the customer requirements, constraints, assumptions, and risks are for your opportunity. 

This is a live document, and questions will be expand and update over time.

Databases

Generic

 1.  Is this replacing a current solution, or is this a net new project?  What’s the current solution?

Why ask? This question helps us understand the use case, any current expectations and what the competitive landscape may look like

2.  Is the current environment coming to the end of a contract and due for contract renewal/hardware refresh?  how soon?

Why ask?  It helps us understand how serious the customer is about migrating and the drivers : usually cost and helps create a pipe-line.

3.  Is the current Infrastructure solution bare-metal/3-tier or virtualized or Engineered Appliance ? provide details
      a.  e.g.  AIX  , Solaris Sparc , VMware Virtualized ,  OVM/KVM, Exadata/ODA (Oracle)
      b.  FC SAN/speed/10GbE Ethernet/iSCSI , Storage Array : Vendor : All Flash/Hybrid
      c.  If possible use automated means to capture configuration and performance information to help with capturing as much information as possible (RVTools, Collector, AWR, LiveOptics, MAP, etc.)

Why ask?  To determine which environment is easier to go after as a starting point.

4.  What are the 3 major pain points in current environment(s)  ( other than end of life/contract).  Examples:
      a.  License Consolidation
      b.  Managing multiple GUIs ( need single pane of Glass)
      c.  Life Cycle Management/Patching
      d.  Performance
      e.  Storage Sprawl due to multiple copies
      f.  Provisioning

Why ask?   Helps us articulate Nutanix Value for Relational Database Workloads.

 5.  How many sites/env.  ( PROD / DR / QA/DEV/Test)

Why ask?   Helps us articulate a Disaster recovery/backup strategy.

6.  How are backups done today : Native or 3rd Party tools , leveraging Snaps/clones?

Why ask?  Whether using third party DR tools ( Zerto/Actifio/SRM) or native database replication.  Whether using third party backup   ( Commvault/VEEAM/VERITAS ) or native tools

7.  Workload types? :  OLTP (Online Transactional Processing) / OLAP (Online Analytical Processing) /DWH (Data Warehouse)

Why ask?   Helps us identify transactional, OLTP, vs Analytical, OLAP/DWH  ( latency sensitivity )

8.  Largest Database size?

Why ask?  Beyond 30 TB , hyperconverged virtualizing may not be beneficial. Need to understand use case

9.  Performance Characteristics  desired : Bandwidth / IOPS / Latency.  These can be given directly from the customer if known, or gathered using local operating system metrics (perfmon/top) or via a discovery tool or script like AWR for Oracle, or a tool like LiveOptics, SolarWinds, etc.

Why ask?  Accurate sizing

 10.  Type of Database Clustering used if Any

Why ask?   Determine if there are potentially any mission critical workloads

MSSQL

SQL Server Inventory Questions:

1.  Number of SQL Server Instances in the environment?

Why ask?  Inventory purposes and Era only supports a single SQL Server instance on the same host.

2.  Number of SQL Server databases in the environment?

Why ask?  Inventory purposes and also helps identify which databases are considered critical for AG (Always On Availability Groups) etc. , databases reside in an instance.

3.  Total size of SQL Server databases in the environment?

Why ask?  Inventory sizing purposes.

4.  SQL Server versions used in the environment?

Why ask?  Different SQL Server versions have different features, limitations etc and also different CU cumulative update levels.  SQL Server stopped issuing service packs in SQL Server 2016 everything now is a CU format. External SQL Server Edition and Version Comparison.

5.  Windows versions used in the environment?

Why ask?  Different Windows versions have different features, limitations and update levels that may affect SQL Server, also driver versions etc.

6.  SQL Server licensing model used in the environment Core, or Server/Cal?

Why ask?  This can help differentiate which licensing  model the customer is using and why.

7.  SQL Server High Availability and Disaster Recovery being used in the environment?   *(Depending on the complexity for HA or DR, this would warrant further discussion with a Database Specialist/Solutions Architect)

Why ask?  This can help determine if shared storage is used such as a SQL Server Failover Cluster Instance (FCI), or a SQL Server Always On Availability Group (AG) which does not require shared storage.  Also is there any multi site replication being used either as a physical storage layer or logical SQL Server layer.

8.  CPU model, type, speed allocated for current/existing SQL Server hosts?

Why ask?  Inventory sizing purposes for baseline.

9.  Number of CPU/Cores allocated for SQL Server hosts?

Why ask?  Inventory sizing purposes for baseline.

10.  Amount of Memory allocated for SQL Server hosts?

Why ask?  Inventory sizing purposes for baseline.

11.  Amount of storage allocated for SQL Server hosts?

Why ask?  Inventory sizing purposes for baseline.

12.  Storage type used for SQL Server hosts, flash, HDD, DAS, SAN etc?

Why ask?  Inventory sizing purposes for baseline helpful in determining expectations with regard to latency.

13.  Network allocation (speed, number of nics) for SQL Server hosts?

Why ask?  Inventory sizing purposes for baseline.

SQL Server Performance Questions:

 1.  What is the total max IOPS required for all SQL Server Instances?

Why ask?  The number of I/O service requests to use as a baseline for their current workload.

2.  What is the latency requirement for SQL Server?

Why ask?  The response time requirement to use as a baseline for their current workload.

3.  What is the bandwidth requirement for SQL Server both read/write?

Why ask?  The throughput requirement to use as a baseline for their current workload.

4.  What is the current SQL Server workload profile read/write ratio?

Why ask?  This helps determine what their workload profile is like and how it will affect our platform (reads are local, writes incur node replication cost) as a baseline.

5.  What is the SQL Server average IO size?

Why ask?  This helps determine what their workload profile I/O size is related to bandwidth .

6.  Top current SQL Server wait statistics during peak workload?

Why ask?  This helps determine what SQL Server is waiting on to process transactions, where there may be a bottleneck.

7.  Current customer Microsoft SQL Server pain points?

Why ask?  This helps narrow the focus and develop a relationship with the customer.  It also assists in focusing on how Nutanix can help alleviate those specific pain points and gives information about how the solution can be shown to resolve those particular pain points.

Oracle

 1.  License Entitlement  ( Cores/NUPS/ELA/ULA/bundled licensing)?

Why ask?  Oracle licensing is expensive and customers want to make the best use of their entitlement when replatforming and not spend more $$ on new licensing when doing a new solution.  Customers are also looking forward to reducing their Oracle License overhead .

2.  Type of Licensing used? :  STD / Enterprise and other options (RAC/Partitioning …etc.).  Each is a paid item.

Why ask?  There may be possibilities to eliminate some Options by using Nutanix Features such as Compression, Encryption, Replication

3.  Is the customer ready to run a “SQL script” or provide details of the environment using RVTools/Collector?

Why ask? When inventorying an Oracle DB environment, you can use the Automatic Workload Repository (AWR) report to gather detailed inventory and performance statistics for an Oracle Database.  Nutanix has an AWR script that can be run to capture the necessary information and is able to be downloaded from within the Sizer Tool.  When adding a Workload select Import, then click the AWR tab and you will see the AWR SQL Script download link.  Once run, you can then upload the output using the Upload File option.

4.  What are the main pain points in the current environment?

5.  When moving to Nutanix would you consider AHV as a hypervisor?

6.  Have you been introduced to Era?

Era

1.  How do you do DB provisioning today and how long does it take to provision a multi-node database cluster?

Why ask :  To find out customer operational efficiency for provisioning . Era can help improve this from weeks to hours.

2.  How many dev/test copies of databases do you have for a your PROD instance(s)?

Why ask : Customers make multiple “full copies” of PROD for test-dev dev/test and use up to 5-10 times the space they need . Era will help in creating space optimized clones of database with “rapid speed”  

3.  What is your typical clone refresh interval and time it takes to refresh a DB clone?

Why ask: Customers using traditional techniques to refresh a copy of a database from a RMAN backup , takes multiple hours and is usually done once a month .  With Era , they can clone everyday or multiple times a day in minutes.

4.  How do you do your Database Patching (Oracle)?

Why ask :  Oracle patching is a huge pain point in large Oracle environments. Era provides a unique way to do “fleet patching”  which will help save 100’s of man hours spent in traditional patching

5.  How do you migrate Databases when required ( Oracle)?

Why ask : Migration is an involved process and a lot of planning and time is required for migration.
Era provides an easy method to “replicate & migrate” databases (Same version) for same-endian formats. ( Linux->Linux or Windows ->Linux)

6.  What is your choice of  Database Replication  (infrastructure/database/hypervisor based)?  Please elaborate.  * (Depending on the complexity of the environment, this would warrant further discussion with a Database Specialist/Solutions Architect)

Why ask:  customers are looking to reduce their software licensing cost of database replication and will look for opportunities to replicate using infrastructure (nutanix replication) . era enables cross-cluster replication including replicating to a NTNX cluster in AWS cloud in an upcoming release 2.0

7.  What are the database engines they currently use?

Posted on

Server Virtualization

What is a Server Virtualization sizing? 

This is the most common workload along with VDI. This can be used for any web app which needs to be sized. Each workload or the application which is to be migrated to the Nutanix software stack is a VM with its own CPU/RAM/Capacity requirements. To simplify for the users, Sizer has set profiles (small,medium,large ) for the VMs but customizable as per the actual application needs.

 

What are profiles in Server Virtualization in Sizer?

Profiles are fixed templates with pre assigned resources in terms of vCPUs, RAM, SSD, HDD to each profile. Broadly, small, medium,large profiles will have different allocation of these resources.

The idea is to facilitate users with the details of a workload (that is a VM)  so they cna quickly fill in number of VMs and Sizer will do the necessary sizing.

Small VM profile template:

Medium VM profile template:

Large VM profile template:

 

What if my VMs are different? Have differen values? 

While these templates and their values are general guidelines, these are customisable.

Clicking on the Customize, opens a  pop-up for user entered values:

Posted on

ECX savings

What is ECX [Erasure Coding] in Nutanix ?

The Nutanix platform leverages a replication factor (RF) for data protection  and availability.  This method provides the highest degree of availability because it does not require reading from more than one storage location or data re-computation on failure.  However, this does come at the cost of storage resources as full copies are required.

To provide a balance between availability while reducing the amount of storage required, DSF provides the ability to encode data using erasure codes (EC).  Similar to the concept of RAID (levels 4, 5, 6, etc.) where parity is calculated, EC encodes a strip of data blocks on different nodes and calculates parity.  In the event of a host and/or disk failure, the parity can be leveraged to calculate any missing data blocks (decoding).

The number of data and parity blocks in a strip is configurable based upon the desired failures to tolerate.  The configuration is commonly referred to as the number of <data blocks>/<number of parity blocks>.

How is ECX savings calculated in Sizer ?

Sizer follows the Nutanix Bible and its guidelines for ECX savings.

Below table shows the ECX overhead vs RF2/RF3 for different nodes:

The expected overhead can be calculated as <# parity blocks> / <# data blocks>.  For example, a 4/1 strip has a 25% overhead or 1.25X compared to the 2X of RF2.  A 4/2 strip has a 50% overhead or 1.5X compared to the 3X of RF3.

 

How does Sizer calculate ECX savings from the above: 

Lets take an example where the cold data for workload is 100TiB.

Also, we will use RF2 as the settings chosen for workload.

So depending on the size of the workload, if the total node recommended came to (lets say 4 nodes), as per the above table: data/parity is 2/1.  So 1.5x overhead for ECX as against 2 for RF2 , thus 50% savings.

For conservative approach and to be on safe side, we only consider ECX for 90 % of the cold data.

ECX applied on 90% of 100TiB = 90TiB

How much ECX savings: 50% = 50% of 90TiB = 45TiB