Analyzing Workstation Storage Performance
In light of the many stories we've written analyzing real-world storage performance, we have enough data to definitively say that, even in intensive tasks, queue depths are generally pretty low.
Although synthetic metrics are great for showing the differences between SSDs on specification sheets, they aren't particularly indicative of what you'll see and "feel" from solid-state storage once it's inside of your chassis.
At the end of the day, the performance benefits attributable to SSDs aren't as inflated as you might think after comparing the speeds and feeds of solid-state and magnetic drives. They're still massively huge, though. In fact, you'll be hard-pressed to switch back to a hard drive after enjoying what a good SSD can do for boot-up and application start times. The video above helps illustrate how much more productive you can be with an SSD. Instead of running a batch file to stress storage performance to the extreme, we decided to emulate real-world use. Consequently, the video represents the real-time activity of compiling code in the background and then moving on to other tasks.
Clearly, when it comes to workstation-class loads, an SSD is going to benefit you (just like it did in our other real-world performance analysis stories). But just because you're using an SSD for work doesn't mean you have the extra budget to pay $1.50 or $2.00 per gigabyte of capacity.
That's why we continue to recommend a tiered storage strategy, employing a large mechanical data disk and a smaller SSD. You throw your operating system on the solid-state drive, along with performance-sensitive applications. Meanwhile, user data goes on the disk. Caching could be another way to help improve the performance of a workstation, without the manual intervention of separate hard drives and SSDs.
More than anything, we're just happy to see all of this real-world data, cumulatively, compiled for the first time. You can bet that we'll be using it in our storage reviews moving forward, backing up our benchmarking methodology.