In many ways, making a good User Interface (UI) seems like a simple task. As evident by so many software makers over the years, it is anything but simple. A good UI looks elegant to the eye, and will become a part of muscle memory without even realizing it. A bad UI can feel like a cruel joke; designed to tease the brain, and frustrate the user. It’s never done intentionally of course. In fact, bad visual and functional designs happen in any industry all the time. Just think of your favorite ugly car. At some point there was an entire committee that gave it a thumbs up. User Experience (UX) design is also an imperfect science, and the impressions are subject to the eyes of the beholder.
A good UI should present function effortlessly. Make the complex simple. However, it is more than just buttons and menus that factor into a user experience. That larger encompassing UX design is what incorporates among other things, functional requirements with a visual interface that is productive and intuitive. PernixData FVP has always received high marks for that user experience. The product not only accelerated storage I/O, but presented itself in such a way that made it informative and desirable to use.
Why the change?
PernixData products (FVP, and the up and coming Architect) now have a standalone, HTML5 interface using your favorite browser. Moving away from the vSphere Web client was a deliberate move that at first impression might be a bit surprising. With changes in needs and expectations comes the challenge of understanding what is the best way to achieve a desired result. Standalone, traditionally compiled clients are not as appealing as they once were for numerous reasons, so adopting a modern web based framework was important.
Moving to a standalone, pure HTML5 UI built from the ground up allowed for these interactions to be built just the way they should be. It removes limits explicitly or implicitly imposed by someone else’s standards. PernixData gets to step away from the shadows of VMware’s current implementation of FLEX. Removing limitations allows for more flexibility now, and in the future.
One of the first impressions that will get will be the performance of the UI. It is quick and snappy. UX pain often begins with performance – whether it is the technical speed, or the ability for a user to find what they want quickly. The new UI continues where the older UI left off; telling more with less, and doing so very quickly.
Looking at the image below, you will also see that the UI was designed for the use with multiple products. The framework is used not only for FVP, but for the upcoming release of PernixData Architect. This allows for transitions between products to be fluid, and intuitive.
New search capabilities
In larger environments, isolating and filtering VMs for deeper review is a valuable feature. Not that big of a deal with a few dozen VMs, but get a few hundred or more VMs, and it becomes difficult to keep track. The quick search abilities allow for real time filtering down of VMs based on search criteria. Highlighting those VMs then allows for easy comparison.
More granularity with the hero numbers
Hero numbers have been a great way to see how much offload has occurred in an infrastructure. How many I/Os offloaded from the Datastore, how much bandwidth never touched your storage infrastructure due to this offload, and how many writes were accelerated. In previous versions, that number started counting from the moment the FVP cluster was created. In FVP 3.0, you get to choose to see how much offload has occurred over a more granular period of time.
New graphs to show cache handling
Previously, the "Hit Rate and Eviction Rate" metric helped express cache usage, and were combined in a single graph. Hit Rate indicated the percentage of reads that were serviced by the acceleration tier. It didn’t measure writes in any way. Eviction Rate indicated the percentage of data that was being evicted from the acceleration tier to make room for new incoming hot data. Each of them now have their own graphs that are more expansive in the information they provide.
As shown below, "Acceleration Rate" is in place of "Hit Rate." This new metric now accounts for both reads and writes. One thing to note is that writes will only show "accelerated" here when in Write Back mode." Even though Write Back and Write Through populate the cache with the same approach, the green "write" line will only indicate acceleration when the VM or VMs are using a Write Back policy.
"Population and Eviction" (as shown below) replaces the latter half of the "Hit Rate and Eviction Rate" metric. Note that Eviction Rate is no longer measured as a percentage, but by actual amount in GB. This is a better way to view it, as the sizes of acceleration tiers vary, and thus the percentage value varied. Now you can tell more accurately how much data is being evicted at any given time. Population rate is exactly as it sounds. This is going to account for write data being placed into the cache regardless of its Write Policy (Write Back or Write Through), as well as data read for the first time from the backing storage, and placed into the cache (known as a "false write"). This graph provides much more detail about how the cache is being utilized in your environment.
Now, if you really want to see some magical charts, and the insights that can be gleaned from them, go take a look at PernixData Architect. I’ll be covering those graphs in more detail in upcoming posts.
A lot of new goodies have been packed into the latest version of FVP, but this covers a bit about why the UI was changed, and how PernixData products are in a great position to evolve and meet the demands of the user and the environment.