How to Do a Virtual SAN Pricing Comparison the Right Way – Part 2

In the first part of How to Do a Virtual SAN Pricing Comparison, we discussed the first 5 steps that include:

• Gathering requirements
• Sizing Virtual SAN
• Determining time horizon
• Picking the right license
• Support inclusion

In this second part I’d like to cover how to use the TCO Calculator to do a customized advanced comparison to any other storage solution, how to calculate OPEX savings when comparing to traditional storage solutions, and what metrics to look at when reviewing the results. Let’s get going then.

Step 6: What do you want to compare to?

Before I dive into details, I’d like to share what I think is an insightful comment provided by our business unit CTO, Christos Karamanolis, after he read the first part of this blog:

“Capacity is only one dimension of the storage system. There is also the performance capability. And with VSAN we give you the ability to size your system for the right balance of performance and capacity capabilities that meet your application needs and your needs for future growth. One can even adapt the ratio over time. With most of today’s systems, you get a fixed level of performance per unit of capacity. So, you may end up over-provisioning and over-paying for capacity you don’t need (or vice versa). Not with VSAN.”

 

This comment is very important in light of what we are discussing in this step. In the TCO calculator we tried to provide a breadth of solutions to compare to:

Note that the above comment especially applies to what most customers use today as a storage solution: storage arrays. Even in all-flash arrays where the performance per GB is very good (yet expensive), it is essentially required to equip such solution with a processing unit that is the equivalent of a super-computer to allow customers to leverage all the flash performance. The reason is simple: storage arrays are a single point of storage service for many compute clusters, all making multiple requests at any given time. With a solution like Virtual SAN, you get a distributed system where requests can be handled by any compute unit in the cluster.

Now that you understand the trade-offs between capacity, performance and different architecture choices, you can pick a desired solution to compare. We only provide Hybrid or all-flash solutions for the above reasons, and because these are the only systems that can provide a reasonable performance and be considered an alternative solution to Virtual SAN (sorry, no low end storage solutions here!).

For this example, I’ll use an entry Hybrid Storage Array. After I picked the solution type, which impacts the overall pricing comparison, I’ll need to answer the following questions:

• Is the alternative solution using Fibre Channel Networking? If so, does Virtual SAN bring savings leveraging 10gbE networking?
• Support cost & period – If you included support for Virtual SAN, what percentage of list price is the alternative solution cost along with a matching support period?
• Usable Capacity – Different solutions have different ratios of usable to raw capacity. Here you can use the same usable capacity you provided in the sizing steps for Virtual SAN or customize it.

If you feel like you want more control on the comparison attributes go ahead and click our “Defaults/Assumptions” button on the top right of the TCO Calculator and behold all the assumptions we are making for alternative solutions:

Here, you can see in the first table there are multiple variables you can change to arrive at your desired comparison. Let’s go over a few of the main columns:

• Base Cost – This would the fixed cost component of your storage solution. A typical storage array even with disks/capacity at all still has a cost, namely the controller, and some basic software components that usually get sold with it. So you can choose to use/change this number or zero this number if you really insist on using a ballpark $/GB price only (next column).
• Additional $/GB – As you can probably guess, this is the variable cost component and should be somewhat of a mixed Flash/HDD $/GB cost that will increase as your capacity and performance requirements increase. If you zeroed out the fixed component, your $/GB is probably higher than the suggested one here (remember, these are list prices).
• Usable/Raw – This number will depend on what RAID (1, 5 or 6) you are using or whether your solution also has deduplication/compression capabilities that increase your usable capacity compared to the raw amount you purchased. In the default table, we assume typical arrays use RAID 5/6 so around 85% of raw capacity is usable, while all flash arrays usually have efficient deduplication and compression running to allow for around 5x usable to raw capacity.
• Max Servers VMs / Max Desktops – This is basically a derivative of the storage solution capacity and performance to host typical server or desktop virtual machines. Once you are sizing for a number bigger than the one provided in the table, the calculator will add another storage array to allow hosting of all the required VMs.
• Max Capacity – This is more relevant in the case of all-flash arrays and other hyper-converged solutions, since regular storage arrays usually have a fairly high capacity limit. As opposed to Virtual SAN, where you have full flexibility in how much capacity per host you want to use, other HCI solutions have a fixed amount of capacity or sometimes 2 options for the same appliance. This means you can’t change your mind later once you picked a certain amount of capacity, and you can’t add disks, even if free disk drive bays exist. In all flash arrays, this will vary: some all-flash arrays have a building block approach, so you’ll need to buy a 2^nd all-flash array to get more capacity, and in some cases you can add a shelf of all flash drives, just like with regular arrays.

The second table below the alternative solutions table will allow you to customize networking costs – both for NICs and Switch, and will adjust it for a per-Port cost for a fair comparison. Note the per-port price is affected by both FC and 10GbE NICs and Switches price points, as well as the number of ports available on the NICs and Switches.

Step 7: What OPEX savings can you expect?

That was one of the toughest questions to answer when we launched Virtual SAN in March 2014. How do you quantify savings in operations?

In order to answer the first part of the expected OPEX savings, we asked a storage analyst team with The Taneja Group to preform lab testing on common set of storage tasks needed to preform over the lifecycle of a storage product, and compare the time/effort it takes on a traditional FC SAN compared to Virtual SAN. To learn more about their research and findings, read Redefining Storage Operations with VMware Virtual SAN, which also includes the full report from The Taneja Group.

If you don’t want to read the full report in order to calculate the expected labor savings, we’ve put the results of that research in the OPEX section in our TCO Calculator:

The main input on your side will be the percent of your IT Storage budget going to labor. The default we use is 26%, which is based on Gartner’s IT Key Metrics report.

If you’d like to further fine-tune the OPEX savings assumptions provided from the research, you can change the VSAN time savings and the portion of each of the main tasks from the overall labor costs.

The second part of the OPEX savings comes from our internal performance team research, who compared the additional power required to run VSAN by adding HDDs & SSDs to servers vs the power required to operate an average storage array. To see the details from the performance test, read Reducing Power Consumption in the vSphere 5.5 Datacenter.

Power consumption also translates to cooling expenses. We used a simple 50% ratio of cooling to power expense in order to factor in cooling savings. If you’d like to adjust any of the assumptions around the power & cooling savings the TCO Calculator lets you edit all of them:

Step 8 (and last!): What’s the right metric to summarize the comparison?

Summarizing the comparison on the total capital expense is very a common practice, and while you should definitely look at what would your capital expense be like with each solution, there are hidden factors you may miss if you focus on just this metric. An important point to mention around the overall capital expense is that a common mistake when comparing traditional storage arrays to Virtual SAN is the inclusion of the server price. Remember: your storage array doesn’t come with free compute (servers), so you either add compute costs on both sides or don’t add it at all. In the TCO Calculator, we will add compute cost when you compare Virtual SAN to other HCI solutions, and only the disks + VSAN software costs (again, vSphere is needed on both sides) when comparing to traditional storage arrays.

The second most popular metric to compare storage solutions is $/GB. We already mentioned why getting to the right $/GB when comparing VSAN to other storage solutions is not easy, but the TCO Calculator does this for you. Also, note that we are using $/Usable GB – as we mentioned in Step 6, depending on the chosen system you can different efficiency from you capacity (most times trading better space efficiency with less performance and vice versa)

A third metric that is very popular when doing desktop virtualization projects is $/VM (or more commonly $/Desktop in such projects). Every storage solution might have an upper bound in terms of the amount of Virtual Machines (or desktops) that it can handle in relation to the implied performance demands. Usually, desktop environments are write heavy and put a lot of load on the chosen storage solution – this where traditional solutions really suffer since the storage controller usually can’t handle more than a certain amount of I/Os per second (IOPS), regardless of the flash capacity on it.

Lastly, maybe a somewhat neglected metric is the $/IOPS – this is the metric that will let you know how much are you paying for ever unit of performance with your chosen storage solution. This where you’ll see Virtual SAN really excel. Virtual SAN leverages the flash tier, along with its native integration in the vSphere kernel to provide exceptional amounts of performance. I’ve yet to see a storage solution that comes close to the $/IOPS efficiencies of Virtual SAN. The engineering team in the VMware storage business unit worked very hard to optimize the schedulers for different workloads that are expected to run on top of vSphere & VSAN and the results are quite impressive – but don’t take it from me, see what the StorageReview team had to say when they tested VSAN: VMware Virtual SAN Review VMmark Performance.

Now that we finished discussing the different metrics, you should also consider the long term implications of choosing one storage solution over the other. Will year 3 capital expenses jump again because you need a refresh? How often you refresh your storage and servers?

The TCO Calculator will bring all this data to light, on both the capital expenses side as well as overall TCO across the selected time period, and will explicitly show you where the savings come from in both numerical summary tables and graphical representation:

Having second thoughts on some of your inputs? Just go back and change them, and everything will be re-calculated and adjusted. Would you like to get back to your analysis later? Don’t forgot to save it! Just click the save button on the top right (you will have to register, so we can associate an account with your analysis). To take your findings and show them to your manager/colleague/customer, click the export to PowerPoint button right next to the save button.

That’s it. I hope you gained some insights on how to do a full blown pricing comparison on Virtual SAN, and if you still have questions, leave them in the comments section below. Thanks for reading!