Here is another great piece re-posted from the OPNET Experts on OPNET's AppResponse Xpert...From the Trenches: "Dear Outsourcer, I have visibility into my own app infrastructure now..You missed something!"
Recently, I was onsite at a major pharmaceutical company. My assignment was to help them use our APM products to solve chronic application performance problems that had plagued multiple key internal applications. Management of most of their key applications and infrastructure were outsourced, and the provider was unable to adequately respond to their complaints. "We didn't write your applications, not our problem." The IT team lacked both visibility and an understanding of the problem to refute such claims. They suspected it was a network problem, however, because so many seemingly unrelated apps were symptomatic. After several weeks of trying with primitive in-house tools, OPNET received a call from the IT team.
It took me 2 hours to set up and configure AppResponse Xpert, our most popular product and also what most of our clients use for "triage". Depending on the initial diagnosis, I was prepared to recommend our other APM products to drill deeper into individual application transactions captured from production servers or the network, but seeing all of the apps together is usually the best first step. AppResponse Xpert was configured to measure end user experience coming into and out of the data center, using live monitoring sessions from SPANs on 4 switches. Using a unique feature for monitoring interface groups, I could immediately see aggregate stats for traffic on each interface combined, as well as from each interface. I took note of which transactions were entering the data center from each WAN switch as well as end-user response times for individual web pages and sites. This led to a brief discussion with the IT department about some interesting observed (and unexpected) patterns of user behavior, but we quickly returned to our investigation..
Our first major clue revealed itself immediately: disproportionately high number of transactions were entering the company's data center on one switch but served on another. Asymmetric routing. I asked the IT staff if this was intended the answer was a resounding "no." Apparently, asymmetric routing was an unintended side effect of the redundant architecture they were sold. The impact of traffic moving across different ISPs should be inconsistent performance, but not necessarily bad performance. This mystery was quickly solved by using AppResponse Xpert to compare end user response times for the symptomatic and well-behaved application sites that had otherwise similar bandwith and latency. The results were conclusive: a particular ISP was not preforming well, so it made no sense to route parts of each app transaction through that ISP.
Does it make sense to leave the productivity of every employee at this company up to chance? I didn't think so, and neither did the IT department. Armed with visibility and knowledge of the root cause, they quickly assigned a specialist to reconfigure the network to consistently route traffic through the right ISPs. Although I didn't stick around to conduct a survey, I'm sure that their employees breathed a sigh of relief when they arrived at work the next day and noticed that their systems were dramatically more responsive. The next time a tidal wave of tickets hits the help desk, hopefully a solution is minutes away vs. weeks now that they have AppResponse Xpert monitoring 24x7.
Find out how to be an OPNET reseller today!