Welcome!

Cloud Expo Authors: Liz McMillan, Pat Romanski, Elizabeth White, Yeshim Deniz, Roger Strukhoff

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We recognize that the average no longer represents the bulk of the transactions but can be a lot higher than the median.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

Cloud Expo Latest Stories
The 16th International Cloud Expo announces that its Call for Papers is now open. 16th International Cloud Expo, to be held June 9–11, 2015, at the Javits Center in New York City brings together Cloud Computing, APM, APIs, Security, Big Data, Internet of Things, DevOps and WebRTC to one location. With cloud computing driving a higher percentage of enterprise IT budgets every year, it becomes increasingly important to plant your flag in this fast-expanding business opportunity. Submit your speaking proposal today!
14th International Cloud Expo, held on June 10–12, 2014 at the Javits Center in New York City, featured three content-packed days with a rich array of sessions about the business and technical value of cloud computing, Internet of Things, Big Data, and DevOps led by exceptional speakers from every sector of the IT ecosystem. The Cloud Expo series is the fastest-growing Enterprise IT event in the past 10 years, devoted to every aspect of delivering massively scalable enterprise IT as a service.
Hardware will never be more valuable than on the day it hits your loading dock. Each day new servers are not deployed to production the business is losing money. While Moore’s Law is typically cited to explain the exponential density growth of chips, a critical consequence of this is rapid depreciation of servers. The hardware for clustered systems (e.g., Hadoop, OpenStack) tends to be significant capital expenses. In his session at 15th Cloud Expo, Mason Katz, CTO and co-founder of StackIQ, to discuss how infrastructure teams should be aware of the capitalization and depreciation model of these expenses to fully understand when and where automation is critical.
Over the last few years the healthcare ecosystem has revolved around innovations in Electronic Health Record (HER) based systems. This evolution has helped us achieve much desired interoperability. Now the focus is shifting to other equally important aspects – scalability and performance. While applying cloud computing environments to the EHR systems, a special consideration needs to be given to the cloud enablement of Veterans Health Information Systems and Technology Architecture (VistA), i.e., the largest single medical system in the United States.
In his session at 15th Cloud Expo, Mark Hinkle, Senior Director, Open Source Solutions at Citrix Systems Inc., will provide overview of the open source software that can be used to deploy and manage a cloud computing environment. He will include information on storage, networking(e.g., OpenDaylight) and compute virtualization (Xen, KVM, LXC) and the orchestration(Apache CloudStack, OpenStack) of the three to build their own cloud services. Speaker Bio: Mark Hinkle is the Senior Director, Open Source Solutions, at Citrix Systems Inc. He joined Citrix as a result of their July 2011 acquisition of Cloud.com where he was their Vice President of Community. He is currently responsible for Citrix open source efforts around the open source cloud computing platform, Apache CloudStack and the Xen Hypervisor. Previously he was the VP of Community at Zenoss Inc., a producer of the open source application, server, and network management software, where he grew the Zenoss Core project to over 10...
Most of today’s hardware manufacturers are building servers with at least one SATA Port, but not every systems engineer utilizes them. This is considered a loss in the game of maximizing potential storage space in a fixed unit. The SATADOM Series was created by Innodisk as a high-performance, small form factor boot drive with low power consumption to be plugged into the unused SATA port on your server board as an alternative to hard drive or USB boot-up. Built for 1U systems, this powerful device is smaller than a one dollar coin, and frees up otherwise dead space on your motherboard. To meet the requirements of tomorrow’s cloud hardware, Innodisk invested internal R&D resources to develop our SATA III series of products. The SATA III SATADOM boasts 500/180MBs R/W Speeds respectively, or double R/W Speed of SATA II products.
As more applications and services move "to the cloud" (public or on-premise) cloud environments are increasingly adopting and building out traditional enterprise features. This in turn is enabling and encouraging cloud adoption from enterprise users. In many ways the definition is blurring as features like continuous operation, geo-distribution or on-demand capacity become the norm. NuoDB is involved in both building enterprise software and using enterprise cloud capabilities. In his session at 15th Cloud Expo, Seth Proctor, CTO at NuoDB, Inc., will discuss the experiences from building, deploying and using enterprise services and suggest some ways to approach moving enterprise applications into a cloud model.
Until recently, many organizations required specialized departments to perform mapping and geospatial analysis, and they used Esri on-premise solutions for that work. In his session at 15th Cloud Expo, Dave Peters, author of the Esri Press book Building a GIS, System Architecture Design Strategies for Managers, will discuss how Esri has successfully included the cloud as a fully integrated SaaS expansion of the ArcGIS mapping platform. Organizations that have incorporated Esri cloud-based applications and content within their business models are reaping huge benefits by directly leveraging cloud-based mapping and analysis capabilities within their existing enterprise investments. The ArcGIS mapping platform includes cloud-based content management and information resources to more widely, efficiently, and affordably deliver real-time actionable information and analysis capabilities to your organization.
Almost everyone sees the potential of Internet of Things but how can businesses truly unlock that potential. The key will be in the ability to discover business insight in the midst of an ocean of Big Data generated from billions of embedded devices via Systems of Discover. Businesses will also need to ensure that they can sustain that insight by leveraging the cloud for global reach, scale and elasticity. In his session at Internet of @ThingsExpo, Mac Devine, Distinguished Engineer at IBM, will discuss bringing these three elements together via Systems of Discover.
Cloud and Big Data present unique dilemmas: embracing the benefits of these new technologies while maintaining the security of your organization’s assets. When an outside party owns, controls and manages your infrastructure and computational resources, how can you be assured that sensitive data remains private and secure? How do you best protect data in mixed use cloud and big data infrastructure sets? Can you still satisfy the full range of reporting, compliance and regulatory requirements? In his session at 15th Cloud Expo, Derek Tumulak, Vice President of Product Management at Vormetric, will discuss how to address data security in cloud and Big Data environments so that your organization isn’t next week’s data breach headline.
The cloud is everywhere and growing, and with it SaaS has become an accepted means for software delivery. SaaS is more than just a technology, it is a thriving business model estimated to be worth around $53 billion dollars by 2015, according to IDC. The question is – how do you build and scale a profitable SaaS business model? In his session at 15th Cloud Expo, Jason Cumberland, Vice President, SaaS Solutions at Dimension Data, will give the audience an understanding of common mistakes businesses make when transitioning to SaaS; how to avoid them; and how to build a profitable and scalable SaaS business.
SYS-CON Events announced today that Gridstore™, the leader in software-defined storage (SDS) purpose-built for Windows Servers and Hyper-V, will exhibit at SYS-CON's 15th International Cloud Expo®, which will take place on November 4–6, 2014, at the Santa Clara Convention Center in Santa Clara, CA. Gridstore™ is the leader in software-defined storage purpose built for virtualization that is designed to accelerate applications in virtualized environments. Using its patented Server-Side Virtual Controller™ Technology (SVCT) to eliminate the I/O blender effect and accelerate applications Gridstore delivers vmOptimized™ Storage that self-optimizes to each application or VM across both virtual and physical environments. Leveraging a grid architecture, Gridstore delivers the first end-to-end storage QoS to ensure the most important App or VM performance is never compromised. The storage grid, that uses Gridstore’s performance optimized nodes or capacity optimized nodes, starts with as few a...
SYS-CON Events announced today that Solgenia, the global market leader in Cloud Collaboration and Cloud Infrastructure software solutions, will exhibit at SYS-CON's 15th International Cloud Expo®, which will take place on November 4–6, 2014, at the Santa Clara Convention Center in Santa Clara, CA. Solgenia is the global market leader in Cloud Collaboration and Cloud Infrastructure software solutions. Designed to “Bridge the Gap” between personal and professional social, mobile and cloud user experiences, our solutions help large and medium-sized organizations dramatically improve productivity, reduce collaboration costs, and increase the overall enterprise value by bringing collaboration and infrastructure solutions to the cloud.
Cloud computing started a technology revolution; now DevOps is driving that revolution forward. By enabling new approaches to service delivery, cloud and DevOps together are delivering even greater speed, agility, and efficiency. No wonder leading innovators are adopting DevOps and cloud together! In his session at DevOps Summit, Andi Mann, Vice President of Strategic Solutions at CA Technologies, will explore the synergies in these two approaches, with practical tips, techniques, research data, war stories, case studies, and recommendations.
Enterprises require the performance, agility and on-demand access of the public cloud, and the management, security and compatibility of the private cloud. The solution? In his session at 15th Cloud Expo, Simone Brunozzi, VP and Chief Technologist(global role) for VMware, will explore how to unlock the power of the hybrid cloud and the steps to get there. He'll discuss the challenges that conventional approaches to both public and private cloud computing, and outline the tough decisions that must be made to accelerate the journey to the hybrid cloud. As part of the transition, an Infrastructure-as-a-Service model will enable enterprise IT to build services beyond their data center while owning what gets moved, when to move it, and for how long. IT can then move forward on what matters most to the organization that it supports – availability, agility and efficiency.