Click here to close now.

Welcome!

Web 2.0 Authors: Plutora Blog, Klaus Enzenhofer, Trevor Parsons, Carmen Gonzalez, Cloud Best Practices Network

Related Topics: Java, SOA & WOA, Adobe Flex, AJAX & REA, Apache

Java: Article

Why Averages Are Inadequate, and Percentiles Are Great

Averages are ineffective because they are too simplistic and one-dimensional

Anyone who ever monitored or analyzed an application uses or has used averages. They are simple to understand and calculate. We tend to ignore just how wrong the picture is that averages paint of the world. To emphasis the point let me give you a real-world example outside of the performance space that I read recently in a newspaper.

The article was explaining that the average salary in a certain region in Europe was 1900 Euro's (to be clear this would be quite good in that region!). However when looking closer they found out that the majority, namely 9 out of 10 people, only earned around 1000 Euros and one would earn 10.000 (I over simplified this of course, but you get the idea). If you do the math you will see that the average of this is indeed 1900, but we can all agree that this does not represent the "average" salary as we would use the word in day to day live. So now let's apply this thinking to application performance.

The Average Response Time
The average response time is by far the most commonly used metric in application performance management. We assume that this represents a "normal" transaction, however this would only be true if the response time is always the same (all transaction run at equal speed) or the response time distribution is roughly bell curved.

A Bell curve represents the "normal" distribution of response times in which the average and the median are the same. It rarely ever occurs in real applications

In a Bell Curve the average (mean) and median are the same. In other words observed performance would represent the majority (half or more than half) of the transactions.

In reality most applications have few very heavy outliers; a statistician would say that the curve has a long tail. A long tail does not imply many slow transactions, but few that are magnitudes slower than the norm.

This is a typical Response Time Distribution with few but heavy outliers - it has a long tail. The average here is dragged to the right by the long tail.

We recognize that the average no longer represents the bulk of the transactions but can be a lot higher than the median.

You can now argue that this is not a problem as long as the average doesn't look better than the median. I would disagree, but let's look at another real-world scenario experienced by many of our customers:

This is another typical Response Time Distribution. Here we have quite a few very fast transactions that drag the average to the left of the actual median

In this case a considerable percentage of transactions are very, very fast (10-20 percent), while the bulk of transactions are several times slower. The median would still tell us the true story, but the average all of a sudden looks a lot faster than most of our transactions actually are. This is very typical in search engines or when caches are involved - some transactions are very fast, but the bulk are normal. Another reason for this scenario are failed transactions, more specifically transactions that failed fast. Many real-world applications have a failure rate of 1-10 percent (due to user errors or validation errors). These failed transactions are often magnitudes faster than the real ones and consequently distorted an average.

Of course performance analysts are not stupid and regularly try to compensate with higher frequency charts (compensating by looking at smaller aggregates visually) and by taking in minimum and maximum observed response times. However we can often only do this if we know the application very well, those unfamiliar with the application might easily misinterpret the charts. Because of the depth and type of knowledge required for this, it's difficult to communicate your analysis to other people - think how many arguments between IT teams have been caused by this. And that's before we even begin to think about communicating with business stakeholders!

A better metric by far are percentiles, because they allow us to understand the distribution. But before we look at percentiles, let's take a look a key feature in every production monitoring solution: Automatic Baselining and Alerting.

Automatic Baselining and Alerting
In real-world environments, performance gets attention when it is poor and has a negative impact on the business and users. But how can we identify performance issues quickly to prevent negative effects? We cannot alert on every slow transaction, since there are always some. In addition, most operations teams have to maintain a large number of applications and are not familiar with all of them, so manually setting thresholds can be inaccurate, quite painful and time-consuming.

The industry has come up with a solution called Automatic Baselining. Baselining calculates out the "normal" performance and only alerts us when an application slows down or produces more errors than usual. Most approaches rely on averages and standard deviations.

Without going into statistical details, this approach again assumes that the response times are distributed over a bell curve:

The Standard Deviation represents 33% of all transactions with the mean as the middle. 2xStandard Deviation represents 66% and thus the majority, everything outside could be considered an outlier. However most real world scenarios are not bell curved...

Typically, transactions that are outside two times standard deviation are treated as slow and captured for analysis. An alert is raised if the average moves significantly. In a bell curve this would account for the slowest 16.5 percent (and you can of course adjust that); however; if the response time distribution does not represent a bell curve, it becomes inaccurate. We either end up with a lot of false positives (transactions that are a lot slower than the average but when looking at the curve lie within the norm) or we miss a lot of problems (false negatives). In addition if the curve is not a bell curve, then the average can differ a lot from the median; applying a standard deviation to such an average can lead to quite a different result than you would expect. To work around this problem these algorithms have many tunable variables and a lot of "hacks" for specific use cases.

Why I Love Percentiles
A percentile tells me which part of the curve I am looking at and how many transactions are represented by that metric. To visualize this look at the following chart:

This chart shows the 50th and 90th percentile along with the average of the same transaction. It shows that the average is influenced far mor heavily by the 90th, thus by outliers and not by the bulk of the transactions

The green line represents the average. As you can see it is very volatile. The other two lines represent the 50th and 90th percentile. As we can see the 50th percentile (or median) is rather stable but has a couple of jumps. These jumps represent real performance degradation for the majority (50%) of the transactions. The 90th percentile (this is the start of the "tail") is a lot more volatile, which means that the outliers slowness depends on data or user behavior. What's important here is that the average is heavily influenced (dragged) by the 90th percentile, the tail, rather than the bulk of the transactions.

If the 50th percentile (median) of a response time is 500ms that means that 50% of my transactions are either as fast or faster than 500ms. If the 90th percentile of the same transaction is at 1000ms it means that 90% are as fast or faster and only 10% are slower. The average in this case could either be lower than 500ms (on a heavy front curve), a lot higher (long tail) or somewhere in between. A percentile gives me a much better sense of my real world performance, because it shows me a slice of my response time curve.

For exactly that reason percentiles are perfect for automatic baselining. If the 50th percentile moves from 500ms to 600ms I know that 50% of my transactions suffered a 20% performance degradation. You need to react to that.

In many cases we see that the 75th or 90th percentile does not change at all in such a scenario. This means the slow transactions didn't get any slower, only the normal ones did. Depending on how long your tail is the average might not have moved at all in such a scenario.!

In other cases we see the 98th percentile degrading from 1s to 1.5 seconds while the 95th is stable at 900ms. This means that your application as a whole is stable, but a few outliers got worse, nothing to worry about immediately. Percentile-based alerts do not suffer from false positives, are a lot less volatile and don't miss any important performance degradations! Consequently a baselining approach that uses percentiles does not require a lot of tuning variables to work effectively.

The screenshot below shows the Median (50th Percentile) for a particular transaction jumping from about 50ms to about 500ms and triggering an alert as it is significantly above the calculated baseline (green line). The chart labeled "Slow Response Time" on the other hand shows the 90th percentile for the same transaction. These "outliers" also show an increase in response time but not significant enough to trigger an alert.

Here we see an automatic baselining dashboard with a violation at the 50th percentile. The violation is quite clear, at the same time the 90th percentile (right upper chart) does not violate. Because the outliers are so much slower than the bulk of the transaction an average would have been influenced by them and would not have have reacted quite as dramatically as the 50th percentile. We might have missed this clear violation!

How Can We Use Percentiles for Tuning?
Percentiles are also great for tuning, and giving your optimizations a particular goal. Let's say that something within my application is too slow in general and I need to make it faster. In this case I want to focus on bringing down the 90th percentile. This would ensure sure that the overall response time of the application goes down. In other cases I have unacceptably long outliers I want to focus on bringing down response time for transactions beyond the 98th or 99th percentile (only outliers). We see a lot of applications that have perfectly acceptable performance for the 90th percentile, with the 98th percentile being magnitudes worse.

In throughput oriented applications on the other hand I would want to make the majority of my transactions very fast, while accepting that an optimization makes a few outliers slower. I might therefore make sure that the 75th percentile goes down while trying to keep the 90th percentile stable or not getting a lot worse.

I could not make the same kind of observations with averages, minimum and maximum, but with percentiles they are very easy indeed.

Conclusion
Averages are ineffective because they are too simplistic and one-dimensional. Percentiles are a really great and easy way of understanding the real performance characteristics of your application. They also provide a great basis for automatic baselining, behavioral learning and optimizing your application with a proper focus. In short, percentiles are great!

More Stories By Michael Kopp

Michael Kopp has over 12 years of experience as an architect and developer in the Enterprise Java space. Before coming to CompuwareAPM dynaTrace he was the Chief Architect at GoldenSource, a major player in the EDM space. In 2009 he joined dynaTrace as a technology strategist in the center of excellence. He specializes application performance management in large scale production environments with special focus on virtualized and cloud environments. His current focus is how to effectively leverage BigData Solutions and how these technologies impact and change the application landscape.

Comments (1) View Comments

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


Most Recent Comments
rtalexander 11/21/12 12:58:00 AM EST

Hey, could you post a reference or two that covers the theory and/or practicalities of the approach you describe?

Thanks!

@ThingsExpo Stories
As organizations shift toward IT-as-a-service models, the need for managing and protecting data residing across physical, virtual, and now cloud environments grows with it. CommVault can ensure protection &E-Discovery of your data – whether in a private cloud, a Service Provider delivered public cloud, or a hybrid cloud environment – across the heterogeneous enterprise. In his session at 16th Cloud Expo, Randy De Meno, Chief Technologist - Windows Products and Microsoft Partnerships, will discuss how to cut costs, scale easily, and unleash insight with CommVault Simpana software, the only si...
Cloud data governance was previously an avoided function when cloud deployments were relatively small. With the rapid adoption in public cloud – both rogue and sanctioned, it’s not uncommon to find regulated data dumped into public cloud and unprotected. This is why enterprises and cloud providers alike need to embrace a cloud data governance function and map policies, processes and technology controls accordingly. In her session at 15th Cloud Expo, Evelyn de Souza, Data Privacy and Compliance Strategy Leader at Cisco Systems, will focus on how to set up a cloud data governance program and s...
Hadoop as a Service (as offered by handful of niche vendors now) is a cloud computing solution that makes medium and large-scale data processing accessible, easy, fast and inexpensive. In his session at Big Data Expo, Kumar Ramamurthy, Vice President and Chief Technologist, EIM & Big Data, at Virtusa, will discuss how this is achieved by eliminating the operational challenges of running Hadoop, so one can focus on business growth. The fragmented Hadoop distribution world and various PaaS solutions that provide a Hadoop flavor either make choices for customers very flexible in the name of opti...
The Workspace-as-a-Service (WaaS) market will grow to $6.4B by 2018. In his session at 16th Cloud Expo, Seth Bostock, CEO of IndependenceIT, will begin by walking the audience through the evolution of Workspace as-a-Service, where it is now vs. where it going. To look beyond the desktop we must understand exactly what WaaS is, who the users are, and where it is going in the future. IT departments, ISVs and service providers must look to workflow and automation capabilities to adapt to growing demand and the rapidly changing workspace model.
Containers and microservices have become topics of intense interest throughout the cloud developer and enterprise IT communities. Accordingly, attendees at the upcoming 16th Cloud Expo at the Javits Center in New York June 9-11 will find fresh new content in a new track called PaaS | Containers & Microservices Containers are not being considered for the first time by the cloud community, but a current era of re-consideration has pushed them to the top of the cloud agenda. With the launch of Docker's initial release in March of 2013, interest was revved up several notches. Then late last...
Roberto Medrano, Executive Vice President at SOA Software, had reached 30,000 page views on his home page - http://RobertoMedrano.SYS-CON.com/ - on the SYS-CON family of online magazines, which includes Cloud Computing Journal, Internet of Things Journal, Big Data Journal, and SOA World Magazine. He is a recognized executive in the information technology fields of SOA, internet security, governance, and compliance. He has extensive experience with both start-ups and large companies, having been involved at the beginning of four IT industries: EDA, Open Systems, Computer Security and now SOA.
HP and Aruba Networks on Monday announced a definitive agreement for HP to acquire Aruba, a provider of next-generation network access solutions for the mobile enterprise, for $24.67 per share in cash. The equity value of the transaction is approximately $3.0 billion, and net of cash and debt approximately $2.7 billion. Both companies' boards of directors have approved the deal. "Enterprises are facing a mobile-first world and are looking for solutions that help them transition legacy investments to the new style of IT," said Meg Whitman, Chairman, President and Chief Executive Officer of HP...
The industrial software market has treated data with the mentality of “collect everything now, worry about how to use it later.” We now find ourselves buried in data, with the pervasive connectivity of the (Industrial) Internet of Things only piling on more numbers. There’s too much data and not enough information. In his session at @ThingsExpo, Bob Gates, Global Marketing Director, GE’s Intelligent Platforms business, to discuss how realizing the power of IoT, software developers are now focused on understanding how industrial data can create intelligence for industrial operations. Imagine ...
Operational Hadoop and the Lambda Architecture for Streaming Data Apache Hadoop is emerging as a distributed platform for handling large and fast incoming streams of data. Predictive maintenance, supply chain optimization, and Internet-of-Things analysis are examples where Hadoop provides the scalable storage, processing, and analytics platform to gain meaningful insights from granular data that is typically only valuable from a large-scale, aggregate view. One architecture useful for capturing and analyzing streaming data is the Lambda Architecture, representing a model of how to analyze rea...
SYS-CON Events announced today that Vitria Technology, Inc. will exhibit at SYS-CON’s @ThingsExpo, which will take place on June 9-11, 2015, at the Javits Center in New York City, NY. Vitria will showcase the company’s new IoT Analytics Platform through live demonstrations at booth #330. Vitria’s IoT Analytics Platform, fully integrated and powered by an operational intelligence engine, enables customers to rapidly build and operationalize advanced analytics to deliver timely business outcomes for use cases across the industrial, enterprise, and consumer segments.
The explosion of connected devices / sensors is creating an ever-expanding set of new and valuable data. In parallel the emerging capability of Big Data technologies to store, access, analyze, and react to this data is producing changes in business models under the umbrella of the Internet of Things (IoT). In particular within the Insurance industry, IoT appears positioned to enable deep changes by altering relationships between insurers, distributors, and the insured. In his session at @ThingsExpo, Michael Sick, a Senior Manager and Big Data Architect within Ernst and Young's Financial Servi...
SYS-CON Events announced today that Open Data Centers (ODC), a carrier-neutral colocation provider, will exhibit at SYS-CON's 16th International Cloud Expo®, which will take place June 9-11, 2015, at the Javits Center in New York City, NY. Open Data Centers is a carrier-neutral data center operator in New Jersey and New York City offering alternative connectivity options for carriers, service providers and enterprise customers.
The explosion of connected devices / sensors is creating an ever-expanding set of new and valuable data. In parallel the emerging capability of Big Data technologies to store, access, analyze, and react to this data is producing changes in business models under the umbrella of the Internet of Things (IoT). In particular within the Insurance industry, IoT appears positioned to enable deep changes by altering relationships between insurers, distributors, and the insured. In his session at @ThingsExpo, Michael Sick, a Senior Manager and Big Data Architect within Ernst and Young's Financial Servi...
PubNub on Monday has announced that it is partnering with IBM to bring its sophisticated real-time data streaming and messaging capabilities to Bluemix, IBM’s cloud development platform. “Today’s app and connected devices require an always-on connection, but building a secure, scalable solution from the ground up is time consuming, resource intensive, and error-prone,” said Todd Greene, CEO of PubNub. “PubNub enables web, mobile and IoT developers building apps on IBM Bluemix to quickly add scalable realtime functionality with minimal effort and cost.”
Sensor-enabled things are becoming more commonplace, precursors to a larger and more complex framework that most consider the ultimate promise of the IoT: things connecting, interacting, sharing, storing, and over time perhaps learning and predicting based on habits, behaviors, location, preferences, purchases and more. In his session at @ThingsExpo, Tom Wesselman, Director of Communications Ecosystem Architecture at Plantronics, will examine the still nascent IoT as it is coalescing, including what it is today, what it might ultimately be, the role of wearable tech, and technology gaps stil...
With several hundred implementations of IoT-enabled solutions in the past 12 months alone, this session will focus on experience over the art of the possible. Many can only imagine the most advanced telematics platform ever deployed, supporting millions of customers, producing tens of thousands events or GBs per trip, and hundreds of TBs per month. With the ability to support a billion sensor events per second, over 30PB of warm data for analytics, and hundreds of PBs for an data analytics archive, in his session at @ThingsExpo, Jim Kaskade, Vice President and General Manager, Big Data & Ana...
In the consumer IoT, everything is new, and the IT world of bits and bytes holds sway. But industrial and commercial realms encompass operational technology (OT) that has been around for 25 or 50 years. This grittier, pre-IP, more hands-on world has much to gain from Industrial IoT (IIoT) applications and principles. But adding sensors and wireless connectivity won’t work in environments that demand unwavering reliability and performance. In his session at @ThingsExpo, Ron Sege, CEO of Echelon, will discuss how as enterprise IT embraces other IoT-related technology trends, enterprises with i...
When it comes to the Internet of Things, hooking up will get you only so far. If you want customers to commit, you need to go beyond simply connecting products. You need to use the devices themselves to transform how you engage with every customer and how you manage the entire product lifecycle. In his session at @ThingsExpo, Sean Lorenz, Technical Product Manager for Xively at LogMeIn, will show how “product relationship management” can help you leverage your connected devices and the data they generate about customer usage and product performance to deliver extremely compelling and reliabl...
The Internet of Things (IoT) is causing data centers to become radically decentralized and atomized within a new paradigm known as “fog computing.” To support IoT applications, such as connected cars and smart grids, data centers' core functions will be decentralized out to the network's edges and endpoints (aka “fogs”). As this trend takes hold, Big Data analytics platforms will focus on high-volume log analysis (aka “logs”) and rely heavily on cognitive-computing algorithms (aka “cogs”) to make sense of it all.
One of the biggest impacts of the Internet of Things is and will continue to be on data; specifically data volume, management and usage. Companies are scrambling to adapt to this new and unpredictable data reality with legacy infrastructure that cannot handle the speed and volume of data. In his session at @ThingsExpo, Don DeLoach, CEO and president of Infobright, will discuss how companies need to rethink their data infrastructure to participate in the IoT, including: Data storage: Understanding the kinds of data: structured, unstructured, big/small? Analytics: What kinds and how responsiv...