Here at True North we have a long-standing relationship with PernixData going back to version 1 of their FVP product. So it’s been a few years of both using their product and watching their product and company grow into the organization it is today. We have been super impressed, not only with their product line but also with their excellent, dedicated, customer focused staff. The core of their FVP line is to make customers’ lives better. They truly allow customers to do more with what they have or can easily source. The PernixData staff is fully aligned with that mission and we have always felt fully supported while tackling our initiatives.
We were recently engaged with a customer operating a legacy VMware stack consisting of 6 year old hosts, non HCL switching and insufficient storage hardware. Despite the non HCL switching, this stack was probably sufficient for the customers needs 6 years ago with a load appropriate to the organization at that time. The customer was under heavy pressure to increase the workload in the legacy stack and push forward with additional services requiring additional load in an already IOPS starved and network constrained environment. Post load increase we were contacted to evaluate the environment and analyze the VMware environment as a whole. Performance of their entire stack had degraded to nearly unusable and application performance was severely degraded. Roll back or scale back of the application adoption was not an option. Additional sites and users had come online in a very aggressive schedule and the problem had to be resolved immediately. We were given a maintenance window of 10 hours to do whatever we needed to “bring this up to speed”. Ten hours is an amazing window to work with but there was no way we could rip/replace the stack in this time frame.
Upon inspection of the stack it was quickly determined that the main issue at hand was located at the storage array infrastructure. The demand on the legacy equipment was just too high. In reality it had been too high for what appears to be many years. Unfortunately, this customer operated under the impression of degraded performance normality. Adding additional load in the last few days pushed the stack to the actual breaking point. While the compute layer was aged passed what it typically considered a usable life span, there was sufficient capacity and no contention for compute resources.
We were left with few options in a 10 hour time frame to make something out of the situation we were confronted with. It was evident to us where the problem existed and we knew we could not properly source and replace the equipment in 10 hours or migrate data. The solution was glaringly obvious to us. We had to install a server side caching solution and we already knew who to turn to. PernixData was the clear and obvious choice.
We knew from experience that the PernixData FVP product would do exactly what we needed in this situation. We knew we could source SSDs in the time frame and the next step was to reach out to PernixData to download their Bits. Our engineers placed a call to PernixData customer service and explained our situation. The customer service representative was eager to assist and provided us with a 30 day license and access to the FVP software.
We also reached out to a PernixData employee with their engineering team who replied immediately after hours with willingness to assist if required.
Our staff was able to install the SSD disks and configure FVP within an hour or so. The process from start to finish is very quick and intuitive. Their documentation is top notch. We have experience with the FVP product but for those who do not, this is no-brainer install. There are architectural things to consider such as SSD size, if you want to cache reads and writes and how many copies of the cache data to store. For the most part, it’s super easy to deploy and their staff is eager to assist.
We ended up not needing the whole maintenance window and given sufficient host compute capacity this solution can be deployed by rolling through hosts in maintenance mode. Downtime is not required.
We immediately saw metrics indicating improved performance under the hood. Latency data started to improve and IOPS stress on the legacy storage array was decreased.
The FVP Dashboard data indicated how much IO and bandwidth we had saved. The data was very impressive for the limited time it had been in production. Between 12am and 6am the savings were 6.8Million IOs saved from the datastores and 215GB of bandwidth to the datastores. And this was during off peak hours. Read caching was at 60%. We only enabled read cache to begin with.
Customer perception and onsite support staff would be the true indication of success. At 9am we received word from multiple sites that the system and applications have never performed better. In the 6 years of production, the system performance was drastically increased within a few hours.
This success allows our client to serve their customers better and that’s what matters to us. That’s what matters to PernixData as well. PernixData empowered us to empower our client. So how do they do it?
The PernixData FVP product allows organizations to offload flash read and write at the host and “decouple” performance from storage arrays. The host cache can consist of SSD, PCIe flash or RAM. Keeping the cache at the host and near the actual compute increases application performance and removes load from the storage network and storage array infrastructure.
Here are some of the lines directly from the PernixData Site
“PernixData FVP software is the premier platform for server-side storage acceleration, delivering faster VM performance than all-flash arrays at a fraction of the cost. Get the fastest virtual machine performance, predictable scale-out growth, push-button simplicity and no rip-and-replace headaches.”
“Increase performance simply by adding more server compute, flash, or RAM. Decoupling performance from array capacity lets you predictably and cost-effectively scale performance as your environment grows.”
The overall outcome in terms of performance was tremendous. Multiple options are now available for this customer to properly migrate their infrastructure correctly. And of course, the recommendation for PernixData for server side caching will be supplied.
PernixData really makes it look easy with their info graphics and in reality it is that simple. It just works really well and their staff is dedicated to making success stories like this one. If we had a cape and came to the rescue it may look like a NASCAR hood with corporate logos. One of the big logos would be PernixData! Thanks for sponsoring our success.