Network Computing published a three-part series on storage performance written by Howard Marks recently. Each of the posts touched on a different topic related to storage performance. The first installment covered metrics and the second part looked at considering IOPS and latency together. The third one discussed how RAID affects performance including reliability and availability.
I’m pointing this series out because the posts are worth a read if you care about storage performance. They serve as a good reminder that performance is not just about speed and reinforce the notion that evaluating performance in context with your applications is critical. Below I have highlighted the discussion in each of the three articles:
- “Storage performance isn’t as simple as bandwidth or latency alone.”
- Evaluating performance on a streaming application is different than doing the same on a database application. One case values throughput at a premium, the other values latency.
This is why when you are considering a SSD SAN storage vendor that you have them either analyze your existing application performance (Kaminario does this for free) or demand your own tests of their system in a non-production environment.
- “A big part of the problem is there’s no such thing as a standard I/O operation.”
- “The key to delivering high application performance is a combination of high IOPS and low latency. As an application or benchmark stresses a storage system, it may continue to deliver high IOPS but at higher levels of latency, and that may seriously affect real-world performance.”
This is why Kaminario has invested in performance tests like the SPC-1 and are continuing to explore additional tests to showcase how the K2 performs in situations that matter to both application and storage managers. You don’t want to measure a car’s performance exclusively by how many miles-per-hour (MPH) it can go. Taking 4 seconds to go from zero to 60MPH is a lot better than taking 12 seconds to go from zero to 60MPH even if with both cars you can go 60MPH. You have to factor latency because it can devalue IOPS.
- “In general, RAID increases reliability and availability of your storage system. The redundancy that it provides comes at a cost, however–not just in additional disk space consumed, but also in the increased amount of work that your disk drives, spinning or solid state, have to do when you write data to a RAID set.”
- The article discusses how “a theoretical RAID controller behaves when reading and writing to some common RAID configurations”
As I have said before, we are in a prove-it business. The good news is that more and more SSD SAN storage discussions are focusing around performance and honing in on what is real versus what is not. I think that is the logical next step now that the market has bought in to the SSD concept. People want apples-to-apples comparisons so they can make the best purchase decisions for their needs. Howard Marks gives buyers good thoughts to consider in this environment.