October 10, 2013, 3:54 PM — It's difficult to identify and address application performance issues when they're tied to storage I/O bottlenecks, but a company that specializes in data analysis has found a way to eliminate those storage performance roadblocks - it hopes once and for all.
Pete Koehler, the IT Manager and Virtualization Architect for Tecplot, says his company was looking for a storage acceleration option that didn't involve buying an entirely new array.
When Storage Bottlenecks Block Innovation
Tecplot, based in Bellvue, Wash., provides data visualization and analysis software for scientists and engineers to help them better understand and analyze data, Koehler says.
The company's software is constantly evolving, but underperforming storage was hampering its developers and slowing down the development cycle, he says.
"In a software development environment, we need to be able to compile code quickly so developers can see the results of their work faster, address bugs more quickly and get the product out to the customers," Koehler says.
"And we had a problem in our environment: the sophistication had grown, but our backend storage couldn't keep up. It wasn't giving us the speed we needed for our developers to work at their full capacity," Koehler says.
Tecplot began looking at the conventional solution to the problem: adding a larger storage array, but quickly turned to more innovative options, Koehler says.
"Historically, we'd shell out a large sum of money to buy a hardware-based solution, and hoped and prayed that was the right choice. Sometimes it worked, and sometimes it didn't," Koehler says.
Even if it did work, "you were stuck with this big, expensive array," Koehler says. "This wasn't something we wanted to do again, so we were looking into options for host-based acceleration so we wouldn't have to add capacity," he says.
Tecplot took a chance on PernixData, and participated in the beta program with about 100 other beta customers. Once the beta was over, Koehler says, Tecplot continued to use the solution and hasn't looked back.
"PernixData FVP does exactly what we need it to do, and works with our existing infrastructure. FVP improves the storage performance on the back end so our developers can do their jobs better and faster," he says.
How PernixData FVP Works
By virtualizing server-side flash into a logical pool, PernixData FVP, released to general availability August 6, decouples storage capacity from storage performance, allowing an increase in speed without adding costly and unnecessary capacity, says Aaron.
"The easiest answer to overcoming the lower latency, speed and performance issues within a virtualized data center used to be adding more capacity," Aaron says. "If you needed more speed, traditionally the answer's been to add more capacity; that's costly, inefficient, and often, only marginally effective," he says.
But by adding flash hypervisor clusters to the server side, Aaron says, PernixData FVP accelerates read-write performance much closer to the application itself, and thus addresses performance bottlenecks where they occur.
"We see customers with maybe 200TB of capacity in their storage arrays, but they're only using 20TB for storage," says co-founder and CEO Poojan Kumar. "The rest is excess capacity they bought to increase performance. That's not the way to solve the problem."
Rather than needing to rip-and-replace or invest in storage arrays, Kumar says, PernixData's customers buy a piece of software, download it and within about 20 minutes, they're up and running on the same hardware and infrastructure they had.
"We tell potential customers, 'Use whatever you like. We just ship software,'" he says. "Add some flash, download the software, make a couple clicks with the mouse, and you're ready to go," Kumar says.
Sharon Florentine covers IT careers and data center topics for CIO.com. Follow Sharon on Twitter @MyShar0na. Email her at firstname.lastname@example.org Follow everything from CIO.com on Twitter @CIOonline and on Facebook.
Read more about architecture in CIO's Architecture Drilldown.