5 minute readPerformance Troubleshooting Series: Solving Performance and Capacity Issues
If you are having performance problems within your storage environment, one of the first things you will need to do is to check the overall health of your array. How do you do that? Look at the performance data. Here are some of the things you can look at in this performance data to assess the overall health of your array:
• SP Utilization: Make sure each storage processor’s utilization is not over 50%. Why 50%? Storage processors are meant to be redundant, in that if one SP was to fail, the other will take over. You want to ensure that the other SP has enough capacity or CPU cycles on it to be able to take the load from the other SP in the event of a failure.
• Cache Utilization: Look at the dirty pages (cache) on the SPs and see if the number is too high. In a storage array, especially on VNXs, you can run into a situation known as forced flushing. Forced flushing is basically when your SP cache is at 100% and your system can no longer accept any more I/Os until it dumps some of the existing cache to disk.
Causes of Performance Bottlenecks
Once you have assessed the overall health of your array, you need to figure out what is causing your performance bottlenecks.
For example, with a high cache utilization, the problem could be because you are writing to the wrong disk types. If you have an environment with high I/O utilization, such as an Oracle database writing to SATA disks, the disks will not be able to keep up with the transactions coming in from your database. Also, if you have a RAIDgroup of, let’s say, 8 disks, you will only have a specific number of IOPs you can actually use. If your database is an OLTP-type database that is constantly hammering to disks, then your RAID group will not be able to keep up with the writes coming in from the host.
Then, your array will take a majority of those writes and write them into cache. This way, the system can keep working, but will fill up the cache quickly. Once the cache has filled up, your array will then go into forced flushing. During this time, your SP will no longer be able to accept anymore writes into cache until it can dump the dirty pages onto the disks. Forced flushing can be detrimental to your system and cause outages.
Questions to Ask Yourself
So, when you’re assessing the overall health of your array you should ask yourself the following:
• Is my SP utilization high? If so, why?
• Is my cache being heavily utilized? If so, why?
• Is my environment being written to the wrong types of disks?
Making Strides to Better Performance
Once you have identified the environment that is having performance issues, you can then make strides to moving it to faster spindles. So, if you use the example of having an Oracle database being written to SATA disks, you can move those data files and lines from a SATA tier to a Fibre Channel tier. This will reduce the cache rate on your SP and will increase the overall performance of your system.
Another option you can use, that is becoming increasingly popular, is to add an EFD tier to your storage array. What is an EFD? An EFD is a solid state disk that acts as another layer of cache for your system. VNX systems have what is called FAST cache that you can add disks into the FAST cache tier that acts as another layer that is able to handle any I/O bursts coming into your system. The EFD tier will give your array some wiggle room that is needed in order to be able to handle an I/O event so that you do not have an array-wide performance issue.
These are just some of the ways to assess the overall health of your storage array and ways to go about remedying the causes of performance issues.If you want to learn more, or want to start making changes to your equipment to increase your performance, contact one of our sales representatives to help you get what you need to empower you to make a difference in your organization.