Performance & Capacity of Storage

I/O Performance Management is about ensuring daily service levels for storage are continually met. This requires an understanding about the performance requirements of your workloads and how your storage hardware has been, is now, and will be handling the workloads.

With automated health checks and quick drill down into root causes, it is now possible to avoid storage performance problems rather than deal with them after they occur. And it is possible to significantly increase the performance levels of existing hardware. See the Storage Performance Management Maturity Matrix for more details about stages of maturity in I/O performance management.

Capacity Planning is the art and science of predicting future workload requirements. An accurate understanding of your workload requirements (historical and future) allows you to correlate those requirements to the capabilities of storage hardware for purchase decisions (Storage Engineering), especially if you use a storage performance modeling package.

Capacity planning as it relates to enterprise storage systems has largely been focused on understanding how much space is and will be required for the data from your applications.

However, disk drives (the backbone of storage systems) actually have two capacity dimensions:

Space capacity - how many GB or TB of data the storage holds.
Performance Capacity - how quickly the data can be accessed in terms of IO's per second.

The term "performance capacity" is just now becoming more used with regards to storage, so it may be unfamiliar to some. Especially because we have trained ourselves for so long to think of storage capacity only as referring to space. But "Performance Capacity" is in the same type of usage as when we talk about the "capacity" of a manufacturing plant or even the capacity of the CPU, for example saying "our CPU peaks at 96% capacity".

The maximum performance capacity of disk drives is dependent upon how fast they spin which is the predominate factor determining the maximum number of read and write IO operations a drive can do.

The space capacity on a drive has doubled many times (36GB, 73GB, 146GB, 300GB, 600GB, etc.) in recent years but the performance capacity of a drive has remained flat (max RPM of 15K for spinning disks). As the drives hold more and more data, they also require more and more back-end IO requests per spindle. Consequently, we are now at the point where performance is a far greater constraint than space.

For most workloads, 300GB drives represent the point where the drives can no longer keep up with the average amount of IO requests for the average amount of data that resides on the drives. This is the primary reason why Solid State Devices (SSDs) are now coming to the forefront as a necessary compliment to large drive configurations.

Space is now cheap and performance is expensive!

Space capacity utilization is fairly easy to track. The amount of space on a drive is a constant number and the amount of space that a GB of data requires also remains constant at one GB. So space capacity utilization is simply the amount of the space available on the physical drives that your data is currently using. For most mainframe environments space capacity utilization is in the range of 60% to 80% which is much higher than space utilization on most distributed systems storage environments.

Tracking performance capacity utilization however, is not so easy and requires specialized software and tools. This is because the performance requirements of each specific GB of data vary widely over different intervals of time. It is also because there are various components inside the storage system that contribute to performance, and visibility into these components is difficult. Consequently, the level of utilization or efficiency of storage performance capacity is not tracked at most sites today.

The IntelliMagic products solve these problems and allow you to do accurate performance capacity planning for your requirements.

IntelliMagic Vision provides you easy visibility into the performance requirements of your workloads on a daily basis from the perspective inside the storage system where today's performance bottlenecks occur. The performance metrics we collect and calculate are stored in a historical trending database so that you can see how these metrics for your workloads have been growing over time. IntelliMagic Vision also provides extensive capabilities to plan and track the bandwidth requirements for both synchronous and asynchronous Copy Services implementations.

IntelliMagic Balance will identify the latent performance capacity of your storage hardware that is caused by imbalance of workloads across the hardware resources you purchased. Very often, there is at least 30% of performance capacity of a storage system that can be added to your available performance capacity by rebalancing of the workloads through some simple logical volume moves once or twice a month. This can provide immediate relief to performance stressed storage systems, and will extend the life of both existing and new hardware. It can also eliminate or reduce the need for expensive SSDs.

IntelliMagic Direction uses the data and knowledge about the performance requirements of your workloads that IntelliMagic Vision produces to show you what specific hardware upgrades or replacements will work best, for example more channels for more bandwidth, or more disk drives (HDDs) to handle the physical disk IOs.

Capacity planning for z/OS Tape Workloads is supported through our Batch Magic product. Batch Magic tracks the number of specific and scratch mounts and the number of MByte/s read and written among many other variables. With this information you can size the bandwidth requirements for both locally attached and remote tape and virtual tape configurations.