Though it looks like a complicated page it is organized neatly into different parts.  Looking  from upper right and going clockwise

• Sizing Options –  This shows current specification on how you want Sizer to do a sizing like Automatic with All Flash, Manual, etc.
• Hardware summary.  Shows the model that was recommended.  Mulntiple rows cover mixed clusters with different types of nodes.
• Sizing Summary.  This shows the current results for the recommendation.  The dials show the utilization for cpu, RAM, HDD, SSD for all clusters combined.    Later Sizer will allow for a per-cluster view
• Sizing Details.  Here all the workloads are summarized and the total required resources are summed for all the workloads.   In the larger table the recommendation’s sizing details are shown in terms of cpu, ram, hdd, ssd usage to cover the workloads, RF2, CVM, etc
• Workloads.  In the left panel are the list of workloads in the scenario
• Actions button.  Here various actions can be performed on the scenario such as downloading a BOM.

Sizing Detail Video

Sizing Details:

First of all you can look at Sizing Details for each cluster or All clusters.

In the top part are the resource requirements  (Cores, RAM, HDD, SSD) from one or more workloads.  These numbers are computed by the workload modules.  Here we have a couple workloads and the resources are totaled under Total

The bottom section is the Capacity calculations for the recommendation.

• The Raw Capacity row is the total available resources (cores, RAM, HDD, SSD) in the cluster(s).    A few things need to be explained
  • Cores is adjusted by specint weight and if there are memory issues like unbalanced DIMMS
    • here is the tooltip which shows the adjustment.  There are total of 72 physical cores in the recommendation.  This is number of models * number of cpu’s per model  * number of cores per cpu
    • Sizer needs to adjust different core speeds as they can vary widely and SpecINt2006 is the standard used.  Here it is -2.23 vs baseline processor.    More information on how Applied Weight is calculated here Sizing Approaches and Logic
    • Sizer does test for unbalanced dimms and here there is no issue and so adjustment = 0.  More information on how Unbalanced DIMMs is accounted for is here Sizing Approaches and Logic

• So for Sizer in this example it assumes there are 69.77 “baseline” cores.

RAM –  Here Sizer computes total RAM and converts it to TiB.

• Here there are 3 nodes and each node has 32GB x 8 Dimms for a total of 768GB of RAM in the recommendation
• A note of trivia, RAM is sold as GB but in reality is GiB.  Since 32GB is less than 32 GiB there is no issue in marketing them in GB as customer is getting more than they expected.  Sizer is very exact though and so we acknowledge its full potential of 32 GiB per DIMM and so here there are 768 GiB in the recommendation
• Converting to TiB it is 768 GiB/1024 = 0.75 TiB  ( there are 1024 GiB in 1 TiB, while 1000 GB in  1 TB)

HDD   Here Sizer computes total HDD and converts it to TiB.

• Here there are 3 nodes and each node has 4x2TB drives  for a total of 24TB of HDD in the recommendation
• Converting to TiB it is 24TB * 0.909494702 TiB/TB = 21.83 TiB

SSD   Here Sizer computes total SSD and converts it to TiB.

• Here there are 3 nodes and each node has 2x960GB drives  for a total of 5760GB of SSD in the recommendation
• We do allocate space for downstroking (80GB per drive) for regular SSDs but no downstroking if Encrypted drives
• so here need to discount downstroking for 6 drives or 480GB.  Now the net is 5760 GB – 480 GB = 5280 GB.  This is 5.28 TB as there are 1000 GB per TB.
• Converting to TiB it is 5.28TB   * 0.909494702 TiB/TB = 4.80 TiB

Compression Savings is the total amount of compression savings found in the workload modules.

Dedupe Savings and ECX Savings (none in this example) are found in the workload modules

RF Overhead is the total amount of HDD or SSD needed to meet the RF2 or RF3 requirements that were set in the workload modules.

CVM overhead is calculated by Sizer following latest rules.    See the next section.  Also  more information on how CVM Sizing is here Sizing Approaches and Logic

Workload total is the same numbers from workload section

Usable Remaining Capacity (Assuming RF2)

• HDD Usable Remaining  Capacity = (Raw + Compression Savings + Dedupe Savings + ECX Savings – RF Overhead – CVM overhead ) / 2
• SSD Usable Remaining  Capacity =  (Raw + Compression Savings + Dedupe Savings + ECX Savings – RF Overhead – CVM overhead + Oplog ) / 2
• Notes:
  • If All Flash,  Compression Savings, Dedupe Savings , ECX Savings, RF Overhead,  and CVM overhead that would be attributed to HDD’s is applied to SSDs
  • For SSD Capacity,Oplog is included as part of CVM overhead for SSDs but also added back into Usable capacity as it is a Write log and so is available for user data.

 

Usable Remaining Capacity (Assuming RF3)

• HDD Usable Remaining  Capacity = (Raw + Compression Savings + Dedupe Savings + ECX Savings – RF Overhead – CVM overhead ) / 3
• SSD Usable Remaining  Capacity =  (Raw + Compression Savings + Dedupe Savings + ECX Savings – RF Overhead – CVM overhead + Oplog ) / 3
• Notes:
  • If All Flash,  Compression Savings, Dedupe Savings , ECX Savings, RF Overhead,  and CVM overhead that would be attributed to HDD’s is applied to SSDs
  • For SSD capacity, Oplog is included as part of CVM overhead for SSDs but also added back into Usable capacity  as it is a Write log and so is available for user data.

Extent Store (Assuming RF2)

• As the tooltip indicates Extent Store is RAW less CVM.  It represents the amount of storage left after CVM .  Put another way it is amount storage available to the customer  before workloads are added.   Here workloads are not included.   The workload RF copies are not included.  Storage efficiencies such as compression or dedupe is not included.   Extent store is a concept in the Nutanix Bible.
• HDD Extent Store (Assuming RF2) = (RAW – CVM)/2 = (21.83 – 2.63)/2 = 9.6 TiB
• SSD Extent Store (Assuming RF2) = (RAW – CVM)/2 = (4.8 – 2.03)/2 =  1.38 TiB

Extent Store (Assuming RF3)

As the tooltip indicates Extent Store is RAW less CVM.  It represents the amount of storage left after CVM .  Put another way it is amount storage available to the customer  before workloads are added.   Here workloads are not included.   The workload RF copies are not included.  Storage efficiencies such as compression or dedupe is not included.   Extent store is a concept in the Nutanix Bible.

HDD Extent Store (Assuming RF3) = (RAW – CVM)/2 = (21.83 – 2.63)/3 =  6.4 TiB

SSD Extent Store (Assuming RF3) = (RAW – CVM)/3 = (4.8 – 2.03)/3 =  0.92 TiB

 

• What are the details on CVM overheads?

  • HDD numbers can be seen by clicking the “I” button
  • SSD numbers can be seen by clicking the “I” button
  • In case of AF all the CVM components are applied to SSD CVM
  • In this example here is the HDD

• In this example here is the SSD

Refer to Sizing Approaches and Logic for information on how these numbers are calculated

The point of Sizing Charts is simply to give exact presentation of the Sizing Details in charts.  Any value in Sizing Charts is reflected in Sizing Details.  Sizing Details being thorough is frankly a table with a lot of numbers.  Sizing Charts then puts it in nice charts if user wants to see it.  Also good to capture in proposals

Separate is Storage Calculator which allows you to enter your own set of nodes and see extent store and a derived effective capacity.  That is NOT tied to the scenario in terms of workloads, recommendation, sizing details.  More info on different page.

Here is the Sizing Details for a Scenario

This is sample scenario that is used to describe the charts.

Here is the Sizing Charts for this scenario

There is an option to view all charts at once.  You can see there is a 1:1 coorespondance between the Sizing Details and the charts for Cores, Ram, HDD, and SSD.  Also shown is the breakout for SSD CVM and HDD CVM.  Maybe a technical customer wants to see the details but in graphical form and this would cover it

Each of these can be looked at individually so you can just look at what interests you

Cores

Here you see the sizing elements for Core.  The tooltip shows the applied weight adjustment and memory adjustments.  The donut shows the CVM, workload requirement and usable remaining cores.

RAM

 

This is the RAM chart.  In this scenario there is 17 TiB of RAM available.  CVM consumes 1 TiB, workload consumes 14.45 TiB and 2.04 TiB remaining

HDD

This shows HDD.  The total amount of HDD space is RAW plus storage efficiency which in this case is just compression.  Dedupe of ECX are two other technologies that save space.   So because of compression savings we actually can deal with a total of 418.49 TiB.  That number is the size of the donut chart.  From there things that consume space would be the workload, RF overhead, and CVM.  Usable remaining HDD then is 122.59 TiB.  In sizing details it is reported as Usable remaining capacity  (Assuming RF2) = 122.59/2 = 61.3 or Usable remaining capacity  (Assuming RF3) = 122.59/3 = 40.86

SSD

This shows SSD.  The total amount of SSD space is RAW plus storage efficiency which in this case is just compression.  Dedupe of ECX are two other technologies that save space.    For SSD (as explained in Sizing details) we do also add back oplog because being a write journal it is user space.  So because of compression savings  and adding back oplog we actually can deal with a total of 302.06 TiB.  That number is the size of the donut chart.  From there things that consume space would be the workload, RF overhead, and CVM.  Usable remaining HDD then is 31.34 TiB.  In sizing details it is reported as Usable remaining capacity  (Assuming RF2) = 31.34/2 = 15.67 or Usable remaining capacity  (Assuming RF3) = 31.34/3 = 10.45

HDD CVM

There is a chart that shows the CVM components on HDD.  In case of All Flash solution all CVM components are stored in SSD

SSD CVM

There is a chart that shows the CVM components on SSD.  In case of All Flash solution all CVM components are stored in SSD.

Questions

What is purpose of Storage Charts?   –  Simply to give a graphical picture of the numbers in Sizing Details.  Here you can see Cores, RAM, HDD, SSD that is consumed and what is available for RF2 or RF3.  It is tied to the scenario.  Thus changing workloads or models will change the charts

What is purpose of Storage Calculator?   –  It is separate from the scenario in terms of workloads and recommendations.  It is intended to allow user to scope amount of storage that is available for given set of nodes.  It answers what is the potential storage available for those nodes.

Do I need the Storage Charts?  –  Since it is 100% duplicate of Sizing Details, not necessarily.  It does give a graphical view though

Well the dials of course.  In our user focus groups everyone loved the dials and yep we kept them.  On left is the node count, power, and rack space.

 

What are the thresholds?

Just hover your mouse over CPU, RAM, HDD ,or SSD and can see it.    However, it is 95% for CPU and RAM, 90% for HDD and 98% for SSD.

re the dials set for N+1?

Depends!!   The dials show the utilization for what you configure the sizing for.  So in auto, there is aggressive  (n+0) , standard  (n+1), conservative (n+2).  In manual,  you are in the driver seat and define the models (ok can only have legitimate parts for that model).  So in manual the dials show you what utilization is available for the models you defined.

 

 

Sizer supports multiple clusters.  Each workload can be assigned to any cluster.  Then in the recommendation you can view results for All Clusters or separately on a cluster by cluster basis

Here is recommendation for All Clusters,  All the hardware and all the workloads are combined in one view.

 

The Sizing Details also reflect the sizing information for all the clusters combined.

Here is same recommendation but for just one cluster.  Note just the  hardware and the workloads assigned to that cluster is in the view.

 

The Sizing Details reflect the sizing information just for that cluster.

Alternatively can look at Rack View for All Clusters or on a cluster by cluster basis

Likewise can view Sizing Charts for All Clusters or on a cluster by cluster basis

Hardware summary shows the model(s) definition of the recommendation.  In a mixed cluster it may have a couple lines for different models.  It also tells you what cluster the models are assigned to.

Ok show me something really cool that I’ll LOVE in Hardware Summary compared to Sizer 2.1?

Click on the “I” and you get the model details and so you DO NOT HAVE to go to MANUAL just to see the model bom  !!!    Ok maybe I’m a nerd but I love it ?

 

What do the letters in the SSD drive indicate?

The letters indicate different levels of endurance in terms of Drive Writes per Day (DWPD).  For example, 3DWPD means you can rewrite all the data on the drive 3 times a day for its entire life that it is warranted for.

Do we support hybrid on non Nutanix models?

Definitely, we support all flash as well as hybrid on all vendors including Lenovo, HPE, Cisco,Dell XC, Dell PE among others..

The Hardware Compatability List [HCL] mentions the hybrid and all flash combinations supported for the different vendors.

Is Sizer aligned with Hardware Compatability List(HCL) for hardware configurations ?

Yes, Sizer, at all times is in sync with HCL on the valid configurtions for across all HW vendors.  At times, certain rules that Sizer implements is mentioned as a note in HCL , so it is advisable to go through the HCL notes to avoid confusion. 

For ex:  In Sizer, for HPE DL380 Gen10 14LFF, using SFF SSD will show invalid configuration unless used in the rear of the server. The same is mentioned in the HPE HCL as a note in the SSD configuration section.  Therefore, faced with any such situation where Sizer shows as invalid configuration, it is advisable to look through HCL notes for clarification. 

On the main scenario page you see the Sizing specification.  Here it is

• Buy
• Homogeneous – All the nodes in the recommendation are IDENTICAL
• Standard Failover
• Any storage is allowed (Hybrid or All Flash)

Automatic Sizing Options

Here you see the current Automatic sizing options

• Buy vs Rent.  For Nutanix there are a few bundles that can be rented.  Default is Buy and most users will leave it there
• Homogeneous vs Mixed – In Homogeneous, all the nodes in the recommendation are IDENTICAL.  This means totally same model with same CPU, RAM, HDD, SSD, NIC.  A mixed cluster allows Sizer to look at second model to come to optimal solution
• Failover
  • Aggressive is N+0.  Sizer finds lowest cost recommendation that meets the workload requirements but does NOT assure N+1 where a node can be taken down (example during upgrade) or lost
  • Standard is N+1.  Sizer finds lowest cost recommendation that meets the workload requirements and adds a node to assure N+1 where a node can be taken down (example during upgrade) or lost
  • Conservative is N+2.  Sizer finds lowest cost recommendation that meets the workload requirements and adds two nodes to assure N+2 where two node can be lost.
  • RECOMMENDATION:  Stay with Standard.  Aggressive might be good for PoC but not in production.  A key value proposition of Nutanix is online upgrades which N+1 allows.  Few users go to Conservative given extra cost of 2 spare nodes
  • Storage
    • Any storage is allowed (Hybrid or All Flash)

All Flash –  Here Sizer will limit recommendations to just All flash models

 

Model Filter

With the model filter can tell Sizer just find the optimal solution for the models you want to focus on.  Say for example your customer has standardized on the NX-3060-G5.  Just click on that model only as shown above and you know Sizer will only look at those models.

 

Manual Sizing options

• You can specify Buy or Rent models.
• On right can add a model or shown below edit a model

We have greatly enhanced our Automatic Sizing.  Auto goes through all the possible options (an option is a model with specific component combination) and finds the optimal solution (lowest net cost)

Certainly good to know and often is the best solution.  We find users want to “play” with the recommendation like

• Increasing and decreasing nodes to see impact on the utilization dials.  Auto provides N+1 or N+2 but good to see impact of losing a node for example
• Would a faster processor or more RAM help a compute intense workload mix?
• How about impact of changing storage options?

In old days you had to go to Manual, make a change, hit apply and then see the impact.  If you want to play with say 10 different changes that is a LOT of effort

Now we have Super Auto !!!

Here is an Automatic Sizing and you see there is 6 nodes.  The one thing that is new is you see the Customize link.  Click on this and you will enter that Super Auto Zone

Here is Super Auto

Key things

• You now have + and – buttons to increment and decrement the nodes and the resources.  In regards to resources
  • Cores – You increment or decrement with more or less overall specint.  We go through the product structure for that model
  • RAM, HDD, SSD – You increment or decrement with more or less capacity.  We go through the product structure for that model
• Whenever you click a new sizing is done with that change in either the model or the node count.  The dials are updated
• You see a cost delta vs optimal.  So you don’t have to go to Budgetary quote to realize relative cost change.  Here we added a node and it increased the expected net price by 15%.  It is approximate.  We have budgetary quote to get a better number
• You see at the top, Sizer tells you it is now customized.  That will be recorded in the BOM too
• Restore to Auto.  Here can just have Sizer go back to optimal.
• Done.  Have fun clicking things and in the end can go with it

What about the product details when I make changes?  

Product component descriptions and quantities are too long to put in the UI by the buttons.  So we have you increment or decrement and you see the overall capacity (e.g. increase or decrease HDD).  However at any time you can look at the model description ( the i by the model) and get all that info