This is the third in a four-part series of posts on the Sun Storage 7000 USS storage arrays. Previous posts in this series can be found here:
In this post, I’ll be discussing some of the analytics features of the 7000 arrays and how the rich data it offers provides customers with added insight into getting the best performance for their money.
Understanding Performance
To many people, storage array performance is a bit of a black art, usually performed by the skilled specialist or vendor and requiring lots of time and effort in planning layout, design and complicated measurement and calculation. On legacy storage arrays this is possibly true, especially where data is tightly located to specific RAID groups or physical disks. Some of the newer storage array technologies (and now features added to older arrays) allows for data to be dispersed across more physical devices and so to gain performance improvements from “wide-striping”. However performances are achieved, there’s one thing for certain, you can’t improve your performance profile unless you can measure it.
USS Analytics
One of the key features of the USS systems is Analytics, which basically translates to rich statistical reporting on the status of the array. Have a look at the graphic on the right here. It shows a section of the dropdown list of metrics that can be selected on the USS for reporting purposes. You can quickly see that reporting breaks down into the basic component functions of the array; CPU, memory, disk and in this instance cache, which relates to the operation of the ZFS filesystem.As an example, I’ve shown a trace of iSCSI and IP throughput for a test client I ran against the test USS, which shows how iSCSI throughput is varying over time. This is a real-time graph from which I’ve extracted a snapshot; for ease of presentation I’ve selected only two metrics but you can add many. What I really like about the USS Analytics is that I can dig down further and look at device performance in more detail. The third thumbnail shows an image listing all of the response times on individual physical disks in the array.
There’s no doubting this is powerful stuff; USS Analytics lets you look at every component of your hardware and software stack, including protocol performance — in detail — both in real time and historically (historical data is kept pretty much forever).Why Analytics Matters
OK, enough of what seems like marketing spin. What counts is how analytics delivers for customers and how it adds value. Well, from my perspective, performance tuning is all about obtaining the data required to make informed changes to the storage environment. USS provides a rich source of performance information, in a simple format and most important — it’s free and bundled into the price. In addition, the data is not being kept in a proprietary format and
can be accessed through CLI and XML interfaces. This is extremely important for users as we move towards the age of commodity storage. Devices that can’t fit into a standard management framework will be at a serious disadvantage as more and more organisations look to deploy storage for the lowest cost without compromising availability and performance.
