Planning for Solid State Drives

IBM i has supported Solid State Drives (SSDs) for quite some time now. The operating system has built-in support for optimizing the use of SSDs as well as commands and parameters that you can use to control data placement on SSDs versus spinning disks. Support for SSDs started with the 5.4 release and has been enhanced in both 6.1 and 7.1, This blog isn't going to go into the details of those capabilities, but rather will talk about how you can determine whether SSDs can provide you with performance advantages.

Not every application will get a performance benefit from SSDs. In general, environments that have read-intensive applications are better candidates to consider for using SSDs; applications that are write-intensive get performance advantages from the write cache in the IOA.

But how can you determine whether your application will get a performance boost from SSDs, sufficient to justify the expense to invest in them?

IBM has made several papers and tools available over the past few years to help you assess and determine whether SSDs can be beneficial. The recommended approach often requires that you have performance skills to make these assessments.

IBM recently made available an offering from the Lab Services Performance and Scalability Services team in Rochester, Minn., to help determine whether your application can benefit from SSDs. This offering from IBM consists of three different options: 

• You can provide the IBM team your Collection Services performance data for review. The IBM i performance experts can get a basic understanding of your I/O performance characteristics from this data to provide some general advice about SSDs.
• You can remotely access systems in Rochester where you can restore your workload/application and do some basic testing to determine the performance advantage with SSDs
• You can chose to do benchmark or performance scalability testing on-site in Rochester with support from IBM i performance experts to optimally configure and exploit SSDs for your application and workloads.

 Of course, the complexity and cost of the effort increases with each option.

The solution brief “IBM i Solid State Drive (SSD) Performance Services” describes the new offering in much more detail and has the necessary contact information if you are interested in pursuing this new offering from IBM.

 

 

 

 

Comments

[Rob Berendt]

Big thing, even when doing the on site, is to make sure you're doing an apples-to-apples comparison. We did such a test of SSD's in an IASP. However due to some difficulties with setting up the journalling we just turned journalling off. That really skewed the results in favor of the SSD's. Money would have better spent on purchasing 5770SS1-42-HA journal performance. That made significantly more of a difference on our machine than SSD's did.

[Rob Berendt]

Would also be nice if
"PTF Requirements for Solid State Drives (SSD)"
http://www-912.ibm.com/s_dir/slkbase.NSF/DocNumber/534676318

was updated and this information available at
"Recommended fixes"
http://www-912.ibm.com/s_dir/slkbase.nsf/recommendedfixes

[Dawn]

Rob,
Your point about "apples-to-apples comparison" is correct; what might seem like minor changes in an environment can indeed skew the results. This is true for any type of performance comparison.

Regarding the second comment about PTFs - I will see what I can do regarding information about the PTFs. If any changes are made, I'll post an update here.

Dawn