Harnessing Big Data Analytics with Riverbed SteelCentral AppResponse 11
Big Data and its challenges
In today’s world where data sets are so large and complex, traditional methods of processing this information are simply inadequate. The challenge to monitoring Big Data is making use of all the information it provides. Every transaction is important, not just the slow ones as you’ll not be able to understand trends without all data points. Many Network Performance Monitoring (NPM) tools summarize performance data by sampling or averaging out data, typically in 1 min intervals. In doing so, they miss the discrete transactions that could be the key to finding trends and patterns that would lead to understanding what the issue(s) are.
The Riverbed solution
Riverbed SteelCentral AppResponse 11 using TruePlot technology can render hundreds of thousands of transactions simultaneously. This level of granularity reveals trends and patterns hidden by traditional NPM tools. TruePlot will not average-out spikes and can clearly differentiate the “symptoms” vs “root causes”.
Let’s look at a situation where the Big Data approach to NPM clearly and simply proves the Load Balancer is causing the application performance issues.
In this use-case, customers were complaining about slow performance from a multi-tier web application. Using SteelCentral AppResponse 11, the SE connected two monitoring points to the appliance. One from the Users perspective “User -> Load Balancer” and the second from the Webservers perspective “Load Balancer -> Webserver”.
TruePlot in action
When looking at the TruePlot diagram from the Users perspective (in front of the load balancer), we are able to see the response time from every transaction in this 1 hour time frame. As you can see, TruePlot shows us every transaction—Big Data in action. The data points are sloping down to the right, meaning that response times start high and decrease over time. This indicates a queuing issue.
Conversely, when looking at the TruePlot diagram from the Webserver perspective (behind the load balancer) response times do not indicate the queuing effect. By having visibility from both perspectives, and having a tool that displays EVERY TRANSACTION—TruePlot, we clearly understand where the problem resides.
By utilizing the power of SteelCentral AppResponse 11 with TruePlot® we have the complete story. With NPM tools that utilize the averages or even percentile method of monitoring Big Data, you can see how they cannot fully represent the behavior like TruePlot does. By measuring aggregates on either side of the Load Balancer, one may be able to deduce there is a problem with the Load Balancer. Using TruePlot technology, the results are irrefutable and further describes the nature of the problem—a queuing issue within the Load Balancer.
Let Riverbed SteelCentral solutions tackle your Big Data analytics and stay out of the War Room!