As promised in my last post, this will be a short blog on which drives should be used for your VSAN deployment. If you have not read my previous blog on VSAN capacity tier guidance, I would recommend you to read it here. This will give you some background and context to this blog post.
Which Drives Should I Choose for Virtual SAN (VSAN) Deployment?
This is the million dollar question. The short answer is: “it depends on your application capacity, availability and performance requirements”. There is no “one size fits all” answer. But I’d like to try and provide some perspective here in helping you make the right decision when choosing your VSAN drives.
Let us take an example for a common use case of a business critical application: Online Transaction Processing (OLTP). Typically, OLTP applications are transaction orientated. That includes order entry, stock level check, financial transactions, customer relationship management, etc. where a large number of users participate in many small transactions. These short transactions need to be fast, and require sub-millisecond response time for each query.
In the application, the technical mapping of these transactions is carried out in 4K or 8K block sizes with 70/30 Read/Write IO profile. This is true for most OLTP applications, but in some cases there could be deviations from this standard profile.
Considerations for an All-Flash VSAN Deployment
In my previous blog I talked about two VSAN deployment options: “Hybrid VSAN” and “All-Flash VSAN”. When we examine OLTP workloads, a “Hybrid VSAN” deployment cannot guarantee sub-millisecond latency consistently for a large number of users over a longer duration. In this case the preferred option is to deploy an “All-Flash VSAN”.
As you examine your All-Flash VSAN you will need to decide which drives to implement in your deployment.
If we look at an OLTP application, the profile is a mix of reads and writes. For these applications, the All-Flash VSAN ‘capacity tier’ will perform the read cache while the VSAN ‘cache tier’ will perform the write cache. (If you are not familiar with VSAN tier and deployment basics, you can learn more about it here.)
In this scenario, when we choose the capacity tier we need to make sure that the drives are resilient enough to accommodate both the “read cache” as well as “data storing” requirements.
This is similar for the cache layer; we need to figure out the size of the cache based on the number of users and how much data they typically write.
VSAN starts de-staging data from cache to capacity after a certain percentage of the cache is filled. If the frequency of data writes is too high then VSAN would de-stage sooner. In that case, we need to make sure that capacity tier drives offer high enough endurance in order to accommodate such frequent writes.
4 Guidelines for Choosing Your Storage Device
I’ve put together a general guideline of the key questions and elements you should examine when choosing your storage drive:
- IO Profile: Is my application read, write or mixed IO profile? Based on application IO nature choose the Virtual SAN Cache tier e.g. in case of OLTP, I would recommend to choose a mixed-use drive type. Mixed use drives are available across SanDisk®’s SSD portfolio, including SAS & SATA, and PCIe Accelerators, and you can learn more about them here.
- Latency: Is my application latency sensitive? When low latency data access is a key priority, you need a solution that can deliver data fast. SSDs such as the SanDisk Lightning Gen. II offer a 12Gb/s SAS interface, and I would also recommend evaluation our Fusion ioMemory PCie Application Accelerators offering a read bandwidth of 2.7GB/s.
- Data set size: How large is my application working data set? Based on the size of your working set you will need to choose the size of cache, and the cache drives’ capacity.
- Endurance: For the capacity tier, medium to low endurance read intensive (RI) drives would fit the bill in most use cases. However, as discussed earlier in this blog, if your application is writing data frequently (such as OLTP applications), you should examine high-endurance drives. Endurance is a specification that does not exist in the hard drive world and understanding SSD endurance is key to ensuring you will see the best ROI from your investment. I suggest reading Hemant Gaidhani’s blog post explaining SSD Endurance and how it affects the life and costs of your flash deployment.
SanDisk has the broadest portfolio of data center flash solutions which encompasses all the above requirements. Based on your actual need, the right drive can be placed in the VSAN solution to make it a successful deployment, meet your application SLA and budget requirements.
I have shared some thoughts and would like to listen to your opinion as well. What have been your considerations when choosing drives for you VSAN deployment? Let me know in the comments below or you can reach me at firstname.lastname@example.org.