The evolution of tech is all about replacing manual tasks with automation, which enables you to perform tasks that were impossible or impractical before. You're suddenly freed from mind-numbing manual configuration -- only to face a fresh load of complexity, as you confront shiny new buttons to push.
That's what has happened to storage. Today we have a vast array of options for host-to-storage connectivity -- each with their own pro/con list and constantly changing best practices -- and an equally wide number of on-array performance and capacity optimization software features. While deploying new storage may take a fraction of the time, knowing how best to deploy it in the first place often requires a lot more thought.
To get the most out of modern enterprise storage, at a minimum you need to stay on top of four areas: monitoring, benchmarking, application characteristics, and a shortlist of general best practices. Some of them are as old as storage itself, while others come courtesy of the increasing complexity to be found in modern enterprise storage tech.
Monitoring
The No. 1 item any storage admin absolutely needs to know is how to monitor storage. Monitoring both capacity and performance will allow you to foresee impending problems before they rise to the level that anyone will notice -- and give you a chance to make necessary configuration changes or put more hardware on order before it's too late. Likewise, monitoring also gives you the data to determine whether or not storage is the cause of an application performance problem.
Most enterprise storage hardware ships with comprehensive monitoring capability, but simply having it set up and running isn't enough. You need to know what you're looking at. Understanding the relationships between throughput, transactional load, transaction size, and read/write mix is critical -- so you can stop problems before you need to diagnose them.
Benchmarking
Benchmarking is the only way to get a feel for a planned workload and determine how the storage you're deploying to meet that load will hold up. That might involve learning to use general storage performance-testing tools like IOMeter, as well as application-specific tools such as JetStress and SQLIO.
Experimenting with benchmarking tools will also give you a much better understanding of how to interpret storage array monitoring data. By creating loads of varying characteristics and observing the impact on your storage, you can also concretely evaluate the impact of various array-side configuration changes.
Knowing your applications
Storage admins can't afford to ignore the application layer. For example, database workloads are among the most demanding you'll encounter. Recently I was faced with a mission-critical database application that had chronic performance issues. At first glance, the problem seemed to be entirely storage-based, with much of the database engine's time spent waiting for storage to cough up data -- but closer examination uncovered design deficiencies within the database itself. Adding a few indices to the database schema to service several frequently used queries cut the storage load by more than 50 percent, and expensive changes to the storage hardware were completely avoided.