Understanding How I/O Workload Profiles Relate to Performance
When implementing a new application there is usually a requirements section or sizing guide that indicates minimum number of processors or cores, the amount of memory, and storage capacity required. More complex application systems such as Microsoft Exchange, SAP, or Oracle E-Business Suite may include sizing calculators to help determine the number of servers, network bandwidth, database requirements, and so on. However, there are relatively few tools when it comes to storage performance sizing, and they tend to be specific to individual products or vendors. Inputs to these tools are often not well understood or simply not known.
As a storage solutions specialist involved in sizing IT infrastructure, Stan Lanning focuses on customer I/O workload profiles that the infrastructure needs to support. He asks his customers a series of questions trying to understand what applications the customer is running or planning to run, how data flows in and out of the system, and hopefully back into several factors that make up the “I/O workload profiles”. Gathering this information is important in order to determine an appropriate performance configuration.
What are these “I/O workload profiles”? Why are they important? How are they identified? How do they relate to performance sizing?
In this Knowledge Sharing article, Stan explains the fundamentals of the characteristics of I/O workload profiles, such as IOPS, MB/s, read/write ratios, response times, and queue lengths. He also explores how different profiles can impact performance sizing and suggests common tools that can be used to identify workload profile characteristics.