Faced with expensive hardware upgrades to improve the performance of its LMS, Alvernia University opted for an I/O-reduction software solution that achieved the same goals at a fraction of the price.
While the need for a fix was undisputed, a hardware upgrade was going to cost upward of $40,000. Instead, the school turned to input/output-reduction software to achieve equivalent performance gains—or better—for 25 percent of the price.
The performance problems faced by Alvernia’s Blackboard system could be traced to I/O inefficiencies in its VWware environment, which—among other problems—was experiencing latencies ranging from 4,000 microseconds to 750,000 microseconds. “We were noticing numerous timeouts, either when students were submitting discussion posts and grades or taking tests,” said Rich Reitenauer, infrastructure management and support manager in Alvernia’s IT department. “I was hearing a lot of complaints.”
In an effort to keep the school’s Blackboard environment stable, Reitenauer had resorted to restarting it every Friday, but this was only a stopgap solution. “We didn’t really want to look at investing in additional solid-state drives, but that was one option,” he recalled. “I started looking at I/O flash cards for VMware, solid-state drives for recache, and a couple of other products.”
Mindful of the department’s budget constraints, though, Reitenauer decided to conduct a trial of Condusiv’s V-locity I/O-reduction software instead.
(Next page: The results of the software trial)
The software works in a virtualized Windows environment to reduce the number of I/Os for a job, so the system can process more data in less time. Virtual environments are prone to a phenomenon known as the I/O blender effect: The workload from multiple virtual machines gets mixed at the hypervisor level, generating a random I/O pattern. Over time, files get broken up into smaller and smaller pieces, resulting in smaller and smaller I/O to process them. “Small, fractured, random I/O really brings system performance to its knees,” said Brian Morin, Condusiv’s senior vice president of global marketing. “V-locity optimizes the I/O profile, so that every I/O carries more data with it. Plus, it’s more sequential in nature so systems can process more data in less time.”
The loss of performance is something that many colleges and universities experience when they virtualize their environments in a bid to wring more efficiency from their servers. At Alvernia, the I/O performance problem did not originate within its Blackboard application, but rather within the SQL database running beneath it. “An application like an LMS writes and retrieves data from a database,” explained Morin. “All the real activity occurs where the data lives, which is in the database. From an I/O standpoint, the application that sits on top of it is just cosmetic.”
For cost reasons, Reitenauer wanted to avoid what he termed a “forklift hardware upgrade” to address the I/O ceiling at Alvernia, but he also knew that the price of an untried software solution would raise red flags. “If I just submitted a purchase request to my boss for that price [about $10,000], I knew it would get denied,” he said.
Fortunately, Condusiv offers a 30-day evaluation of the V-locity software that allowed Reitenauer to build a business case complete with demonstrated performance improvements before the university paid for the product.
According to Reitenauer, the software itself took only minutes to install, but he advises schools to budget at least three days for testing to give the software a chance to optimize performance using self-learning algorithms. A benefit analyzer, which comes bundled with the software, allows users to measure system performance both before and after implementation.
After installing V-locity, Alvernia experienced an enormous improvement in Blackboard’s operation. “We were able to double performance,” said Reitenauer, noting that the amount of time needed to process 1 GB of data dropped from three minutes to 90 seconds. In addition, response times fell from 55 milliseconds to just five.
Patching time for Blackboard itself also significantly decreased, and server startup time for the Blackboard environment went from a range of 7-12 minutes down to 2-3 minutes. “We haven’t had to restart the Blackboard servers for about eight months besides the occasional Windows updates,” said Reitenauer.
In fact, the university was recently able to reduce the number of application servers from four to two, which was a real boon to Reitenauer whose department consists solely of a system network administrator and himself. “Implementing the I/O-reduction software allowed us to work on other projects and cut down our Blackboard administrative time by 85 percent–90 percent,” he said.
But the real test of the product was how campus constituents—students and faculty—reacted. “Everyone is so much happier,” said Reitenauer. “I haven’t had one complaint about Blackboard performance since we implemented V-locity.”
Based on Alvernia’s success with its Blackboard implementation, the university has deployed V-locity with other applications, including its network-monitoring software. By Reitenauer’s calculation, total savings have amounted to $250,000-$300,000—about 75 percent of the school’s capital expenditure budget for an entire year.
“None of this would have happened without that free trial,” said Reitenauer. “It allowed me to put together a business case that proved to my boss that we needed this before we actually purchased it.”
Andrew Barbour is a contributing editor for eCampus News.
- 25 education trends for 2018 - January 1, 2018
- IT #1: 6 essential technologies on the higher ed horizon - December 27, 2017
- #3: 3 big ways today’s college students are different from just a decade ago - December 27, 2017