More APM Data Really Does Beat Better Algorithms

Aternity January 28, 2016


  ‘More data usually beats better algorithms.’  -Anand Rajaraman

The Riverbed approach to big data APM

The debate over application performance management (APM) collection methods continues. Times have changed.  Riverbed has changed while many of our competitors have not. While our competition touts the questionable benefits of old school sampling and snapshotting followed by the use of sophisticated analytics, Riverbed provides today’s APM customers high definition performance data by collecting detail on every transaction in production with minimal overhead. We empower our customers to make better decisions for their IT environment, for their business and ultimately for their bottom line.

To make our point, let’s say that each word in the above paragraph is a transaction. Let’s ‘sample’ the paragraph above by collecting detail on every fourth word. This would be like collecting detail on every fourth transaction that flows through our environment. According to our competition, this will still give us a good understanding of paragraph’s meaning:

Now, let’s take ‘snapshots’ when we see words greater than four characters.  This is much like snapshotting performance when a transaction exceeds a four second threshold:

All three examples have the same word count, the same number of characters and the same word length, but our comprehension of the three are very different.  Of the two ‘low definition’ examples, snapshotting is better than sampling, but we still do not have the complete meaning of the paragraph.  ‘debate about ‘  ‘their bottom ‘  These are important questions answered only by high definition performance data.

Using high definition APM, we get the complete picture and true understanding of our application environment. When we have this, we know without a doubt where to focus our improvement efforts.

In this use case, let’s compare high definition APM with sampling. How does the view change?

This is a screenshot showing high definition metrics based on collecting data on each transaction

If we snapshot and only collect detail on the transactions greater than four seconds over the same timeframe, our view of the application changes.

Here is a quick comparison of the data from the two samples:

High definition

Only transactions greater than 4 sec

Top transaction type by count



Top Transaction by response time



Most active user

Larry Petty

Chong Zhao

User with highest avg response time

Joe McElroy

Frank Ricard

Top URL path by response time



When we have the complete picture of the application, our view of application performance changes. Without the complete picture, you cannot focus on the right issue.  If we only look at the transactions greater than four seconds, we see where Home is the slowest page, but if we look at all transactions, Portfolio is the slowest on average. Our view changes when we have more data with which to make a decision.

The high definition that comes from big data in APM is therefore the only way to truly understand every single issue that impacts an IT environment and its business applications. Big data provides the full picture, helping organizations move from reactive firefighting situations to proactively managing application performance and providing value to the business they support.

Riverbed’s AppInternals is your high definition APM solution. It was engineered from the ground up to be simple and powerful capturing the detail you need to make the right performance and business decision. Visit to learn more and to sign up for the ability to try these capabilities for yourself.