Welcome!

@CloudExpo Authors: Liz McMillan, Yeshim Deniz, PagerDuty Blog, XebiaLabs Blog, Pat Romanski

Related Topics: Java IoT, Microservices Expo, Adobe Flex, Machine Learning , Apache

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We recognize that the average no longer represents the bulk of the transactions but can be a lot higher than the median.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

@CloudExpo Stories
Interoute has announced the integration of its Global Cloud Infrastructure platform with Rancher Labs’ container management platform, Rancher. This approach enables enterprises to accelerate their digital transformation and infrastructure investments. Matthew Finnie, Interoute CTO commented “Enterprises developing and building apps in the cloud and those on a path to Digital Transformation need Digital ICT Infrastructure that allows them to build, test and deploy faster than ever before. The int...
Without a clear strategy for cost control and an architecture designed with cloud services in mind, costs and operational performance can quickly get out of control. To avoid multiple architectural redesigns requires extensive thought and planning. Boundary (now part of BMC) launched a new public-facing multi-tenant high resolution monitoring service on Amazon AWS two years ago, facing challenges and learning best practices in the early days of the new service.
Niagara Networks exhibited at the 19th International Cloud Expo, which took place at the Santa Clara Convention Center in Santa Clara, CA, in November 2016. Niagara Networks offers the highest port-density systems, and the most complete Next-Generation Network Visibility systems including Network Packet Brokers, Bypass Switches, and Network TAPs.
SYS-CON Events announced today that Outlyer, a monitoring service for DevOps and operations teams, has been named “Bronze Sponsor” of SYS-CON's 20th International Cloud Expo®, which will take place on June 6-8, 2017, at the Javits Center in New York City, NY. Outlyer is a monitoring service for DevOps and Operations teams running Cloud, SaaS, Microservices and IoT deployments. Designed for today's dynamic environments that need beyond cloud-scale monitoring, we make monitoring effortless so you ...
My team embarked on building a data lake for our sales and marketing data to better understand customer journeys. This required building a hybrid data pipeline to connect our cloud CRM with the new Hadoop Data Lake. One challenge is that IT was not in a position to provide support until we proved value and marketing did not have the experience, so we embarked on the journey ourselves within the product marketing team for our line of business within Progress. In his session at @BigDataExpo, Sum...
Keeping pace with advancements in software delivery processes and tooling is taxing even for the most proficient organizations. Point tools, platforms, open source and the increasing adoption of private and public cloud services requires strong engineering rigor - all in the face of developer demands to use the tools of choice. As Agile has settled in as a mainstream practice, now DevOps has emerged as the next wave to improve software delivery speed and output. To make DevOps work, organization...
DevOps is often described as a combination of technology and culture. Without both, DevOps isn't complete. However, applying the culture to outdated technology is a recipe for disaster; as response times grow and connections between teams are delayed by technology, the culture will die. A Nutanix Enterprise Cloud has many benefits that provide the needed base for a true DevOps paradigm.
China Unicom exhibit at the 19th International Cloud Expo, which took place at the Santa Clara Convention Center in Santa Clara, CA, in November 2016. China United Network Communications Group Co. Ltd ("China Unicom") was officially established in 2009 on the basis of the merger of former China Netcom and former China Unicom. China Unicom mainly operates a full range of telecommunications services including mobile broadband (GSM, WCDMA, LTE FDD, TD-LTE), fixed-line broadband, ICT, data communica...
As software becomes more and more complex, we, as software developers, have been splitting up our code into smaller and smaller components. This is also true for the environment in which we run our code: going from bare metal, to VMs to the modern-day Cloud Native world of containers, schedulers and micro services. While we have figured out how to run containerized applications in the cloud using schedulers, we've yet to come up with a good solution to bridge the gap between getting your contain...
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, represent...
As organizations realize the scope of the Internet of Things, gaining key insights from Big Data, through the use of advanced analytics, becomes crucial. However, IoT also creates the need for petabyte scale storage of data from millions of devices. A new type of Storage is required which seamlessly integrates robust data analytics with massive scale. These storage systems will act as “smart systems” provide in-place analytics that speed discovery and enable businesses to quickly derive meaningf...
Your homes and cars can be automated and self-serviced. Why can't your storage? From simply asking questions to analyze and troubleshoot your infrastructure, to provisioning storage with snapshots, recovery and replication, your wildest sci-fi dream has come true. In his session at @DevOpsSummit at 20th Cloud Expo, Dan Florea, Director of Product Management at Tintri, will provide a ChatOps demo where you can talk to your storage and manage it from anywhere, through Slack and similar services ...
SYS-CON Events announced today that Ocean9will exhibit at SYS-CON's 20th International Cloud Expo®, which will take place on June 6-8, 2017, at the Javits Center in New York City, NY. Ocean9 provides cloud services for Backup, Disaster Recovery (DRaaS) and instant Innovation, and redefines enterprise infrastructure with its cloud native subscription offerings for mission critical SAP workloads.
Building a cross-cloud operational model can be a daunting task. Per-cloud silos are not the answer, but neither is a fully generic abstraction plane that strips out capabilities unique to a particular provider. In his session at 20th Cloud Expo, Chris Wolf, VP & Chief Technology Officer, Global Field & Industry at VMware, will discuss how successful organizations approach cloud operations and management, with insights into where operations should be centralized and when it’s best to decentraliz...
The taxi industry never saw Uber coming. Startups are a threat to incumbents like never before, and a major enabler for startups is that they are instantly “cloud ready.” If innovation moves at the pace of IT, then your company is in trouble. Why? Because your data center will not keep up with frenetic pace AWS, Microsoft and Google are rolling out new capabilities In his session at 20th Cloud Expo, Don Browning, VP of Cloud Architecture at Turner, will posit that disruption is inevitable for c...
DevOps is often described as a combination of technology and culture. Without both, DevOps isn't complete. However, applying the culture to outdated technology is a recipe for disaster; as response times grow and connections between teams are delayed by technology, the culture will die. A Nutanix Enterprise Cloud has many benefits that provide the needed base for a true DevOps paradigm. In his Day 3 Keynote at 20th Cloud Expo, Chris Brown, a Solutions Marketing Manager at Nutanix, will explore t...
DevOps has often been described in terms of CAMS: Culture, Automation, Measuring, Sharing. While we’ve seen a lot of focus on the “A” and even on the “M”, there are very few examples of why the “C" is equally important in the DevOps equation. In her session at @DevOps Summit, Lori MacVittie, of F5 Networks, explored HTTP/1 and HTTP/2 along with Microservices to illustrate why a collaborative culture between Dev, Ops, and the Network is critical to ensuring success.
SYS-CON Events announced today that SoftLayer, an IBM Company, has been named “Gold Sponsor” of SYS-CON's 18th Cloud Expo, which will take place on June 7-9, 2016, at the Javits Center in New York, New York. SoftLayer, an IBM Company, provides cloud infrastructure as a service from a growing number of data centers and network points of presence around the world. SoftLayer’s customers range from Web startups to global enterprises.
SYS-CON Events announced today that Conference Guru has been named “Media Sponsor” of SYS-CON's 20th International Cloud Expo, which will take place on June 6–8, 2017, at the Javits Center in New York City, NY. A valuable conference experience generates new contacts, sales leads, potential strategic partners and potential investors; helps gather competitive intelligence and even provides inspiration for new products and services. Conference Guru works with conference organizers to pass great dea...
SYS-CON Events announced today that Technologic Systems Inc., an embedded systems solutions company, will exhibit at SYS-CON's @ThingsExpo, which will take place on June 6-8, 2017, at the Javits Center in New York City, NY. Technologic Systems is an embedded systems company with headquarters in Fountain Hills, Arizona. They have been in business for 32 years, helping more than 8,000 OEM customers and building over a hundred COTS products that have never been discontinued. Technologic Systems’ pr...