The concept of I/O wait is very important, and, unfortunately, many application owners and DBAs do not completely understand what it is or how to measure it. This is probably the most important concept to understand when evaluating the I/O performance of your storage. Since I/O wait will be mentioned in many of the future topics to be discussed in the I/O Storm blog, I’ll offer this definition and explanation.

I/O wait is the total time that working processes are blocked, waiting for the I/O operation to complete. Do not confuse I/O wait with the duration time of I/O operations. An I/O operation is a request to retrieve or write data to the storage. The total time that it takes to complete all I/O operations is I/O operation time. To calculate the I/O wait, we need to understand how much time the application was blocked, waiting for all I/O operation to complete. Let’s look at a simple Oracle example (the same scenario will work in SQL Server, as well): Read the rest of this entry »

When discussing storage performance of many RDBMS applications, I/O latency is one of the most important topics. Today, I would like to discuss my view on latency monitoring. In a future post, I will be talking about latency, throughput, and IOPS, and in which situation each is most important.

I/O latency is defined simply as the time that it takes to complete a single I/O operation.

So, what is an acceptable latency number? Before answering this question, I should explain a bit about storage vendors’ published latency numbers. There are two important pieces of information that many vendors neglect to mention: Read the rest of this entry »