This blog post is the second in our series of blog posts of VSAN Evaluation using our integrated sizing and TCO tool ( vsantco.vmware.com ). For part 1 that discusses VM profiling go here:
VSAN Evaluation – How to use the VSAN sizing tool – Part 1
In this part of the blog series we will discuss how to adjust our default recommended ready nodes profiles to your specific scenario. I’m aware this part can be somewhat confusing, but after a few tries I’m sure you’ll get the hang of it and can do this in a matter of minutes.
First, let’s start by explaining what a Virtual SAN Ready Node is. Virtual SAN is a pure software product, but using it with non-certified hardware can have some unwanted results. This is why we decided to partner with all the major OEMs to create ready-to-go hardware solutions called ready nodes that can also be used as a reference architecture and be customized to your preferences. My team member and VSAN Product Manager, Rakesh Radhakrishnan, wrote a blog ( Virtual SAN Ready Nodes – Ready, Set, Go!) almost a year ago answering most common questions around the topic. Also check out the most updated list of ready nodes here – Virtual SAN Ready Nodes – and our Virtual SAN Hardware Quick Reference Guide .
So now that we understand what is a ready node let’s see how we use it in our sizing tool.In the second step of sizing your environment you’ll have a set of ready node profiles around the virtualization platform you picked (server or desktop).
So, if you picked server virtualization for example you’ll get the following options:
Try going back and pick Desktop Virtualization and see how the choices change to the respective linked/full clones ready nodes.
The low-medium-high profiles were built to allow for more performance and more capacity as you go from low to high, but you can really start with any of these and customize as you like – your sizing results will adjust around your choices.
Now let’s get to the customization part and understand the table layout:
In the middle blue column you can see the specs of our selected ready node. In the case above it is the default Hybrid Medium one. If you look again at the Virtual SAN Hardware Quick Reference Guide (under hybrid) you’ll see the matching profile and configuration. The second left column “Modeled Scenario” is what the calculator will use eventually to determine the sizing results and in order to edit that column you’ll need to input changes in the “Build Your Own” column.
So, now you ask what is the most right column, “Optimized for Minimum #HDDs & Cluster suggestion” column, used for? It is there to make your life easy when trying to minimize the hardware you need to build your VSAN cluster. The calculator uses a recursive algorithm to come up with the minimum amount of caching and persistent disks you need to put in each server, in order to get the minimal cluster size – given your selected persistent disk size (1TB, 2TB, 1.2TB, etc.), your desired consolidation ratio (20, 25, 40 VMs per host or even 100 if we are in a desktop scenario – CPU/Memory based), and how many physical drive bays you have in the server.
Sound complicated, but you don’t really need to think about it besides making sure you put your desired environment numbers in. Once you do that you can see what is the minimum amount of disks you need for each server and the minimum number of servers in the cluster.
In our displayed scenario you can see that even though the ready node comes with 2 disk groups of 4 disks each, the calculator suggests we can satisfy our requirements for the VSAN cluster with 1 disk group and 6 disks behind it. Note changing any of the 3 inputs above (number of drive bays in server, VMs/Node, Persistent Disk Type) will impact the optimization calculation.
Pay attention that the optimization is for minimum amount of hardware, and that can impact performance to a certain extent. For example there’s a tradeoff between going with one disk group vs. two ( read Chuck Hollis blog – How to Double your VSAN Performance ), but note that you can always add a 2nd disk group later if you don’t see the need for an increased performance requirement today (yes, you can grow-as-you-go with performance too not just capacity). To understand the IOPS impact you can easily change from 1 to 2 disk groups and back and look how the sizing results change.
If I now change to the minimum optimized 1 flash device and 6 HDDs, a small warning sign pops up:
This means that my Flash/Usable capacity ratio is below the recommended 10%. So, now it means I have two options – either I go back to 2 disk groups or I increase my flash device size in order to comply with our best practice of flash cache sizing. In this example I decided to go with the second option and this is how my final screen looks like:
I no longer have the warning sign and I can continue to the last phase in the sizing process reviewing my results:
I suggest you pay attention specifically to the disk space usage distribution graph – here you can see a fairly accurate picture of the disk capacity usage for the different components – VMDKs, Replicas, Swap file space and additional free capacity. Note, we are also showing you a simulation of the capacity you won’t have available in case of a host failure. The calculator accounts for that in the calculations in the first step (VM Profiling) and makes sure you’ll have enough capacity to fail over to.
Not happy with what you see?
Want more free capacity? go back and add a few more disks to every server or choose larger disks for the persistent tier
Want more performance? go back and add a second disk group or choose faster HDDs
In the next blog in this series I’ll cover how to use your sizing results to create a TCO comparison between VSAN and any other storage solution you are considering.
Links to other blog posts in the series:
Pingback: VSAN Evaluation - How to use the sizing tool - Part 1 | Virtual Blocks - VMware Blogs
Great article, and very useful.
One potential error I have seen on the calculator however is the calculation of the SSD cache/staging disk. This is still at the VSAN 1.0 calculation of 1:10 of usable capacity, whereas this has now changed in VSAN 6 to 10 percent of the anticipated consumed storage capacity – as outlined in the VSAN 6 Design and Sizing guide – and here in an article by Duncan Epping – http://www.yellow-bricks.com/2014/03/07/virtual-san-ga-update-flash-vs-magnetic-disk-ratio/
Don’t get me wrong, I think that the 1:10 rule is a more sensible approach as I would prefer to have more SSD than an exact or too little amount which puts constraints on the VSAN’s ability to grow.
I just wondered if this would be adjusted in a future update?
Thanks for reading the blog.
Regarding your comment – the calculator is using the right calculation. The blog published by Duncan is an old post that was before we changed the guidance to 10% of usable capacity (we were initially more conservative – the blog dates to around our first version launch date).
The design and sizing document specifies as you said “anticipated consumed” which is another way of saying usable. Just take a look at the example following that paragraph (page 23) – the usable capacity is 20TB while the raw is 40TB (ftt=1), but the cache recommendation stays at 10% of 20TB = 2TB.
Hope this clarifies any confusion. Please let me know if I can help any further.