|By Michael Kopp||
|December 12, 2012 07:30 AM EST||
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We recognize that the average no longer represents the bulk of the transactions but can be a lot higher than the median.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.
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!
|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?
SYS-CON Events announced today that Pythian, a global IT services company specializing in helping companies leverage disruptive technologies to optimize revenue-generating systems, has been named “Bronze Sponsor” of SYS-CON's 17th Cloud Expo, which will take place on November 3–5, 2015, at the Santa Clara Convention Center in Santa Clara, CA. Founded in 1997, Pythian is a global IT services company that helps companies compete by adopting disruptive technologies such as cloud, Big Data, advance...
Aug. 29, 2015 11:00 AM EDT Reads: 217
What does “big enough” mean? It’s sometimes useful to argue by reductio ad absurdum. Hello, world doesn’t need to be broken down into smaller services. At the other extreme, building a monolithic enterprise resource planning (ERP) system is just asking for trouble: it’s too big, and it needs to be decomposed.
Aug. 29, 2015 10:00 AM EDT Reads: 326
Early in my DevOps Journey, I was introduced to a book of great significance circulating within the Web Operations industry titled The Phoenix Project. (You can read our review of Gene’s book, if interested.) Written as a novel and loosely based on many of the same principles explored in The Goal, this book has been read and referenced by many who have adopted DevOps into their continuous improvement and software delivery processes around the world. As I began planning my travel schedule last...
Aug. 29, 2015 10:00 AM EDT Reads: 504
Docker containerization is increasingly being used in production environments. How can these environments best be monitored? Monitoring Docker containers as if they are lightweight virtual machines (i.e., monitoring the host from within the container), with all the common metrics that can be captured from an operating system, is an insufficient approach. Docker containers can’t be treated as lightweight virtual machines; they must be treated as what they are: isolated processes running on hosts....
Aug. 29, 2015 10:00 AM EDT Reads: 108
SYS-CON Events announced today that HPM Networks will exhibit at the 17th International Cloud Expo®, which will take place on November 3–5, 2015, at the Santa Clara Convention Center in Santa Clara, CA. For 20 years, HPM Networks has been integrating technology solutions that solve complex business challenges. HPM Networks has designed solutions for both SMB and enterprise customers throughout the San Francisco Bay Area.
Aug. 29, 2015 09:30 AM EDT Reads: 839
SYS-CON Events announced today that G2G3 will exhibit at SYS-CON's @DevOpsSummit Silicon Valley, which will take place on November 3–5, 2015, at the Santa Clara Convention Center in Santa Clara, CA. Based on a collective appreciation for user experience, design, and technology, G2G3 is uniquely qualified and motivated to redefine how organizations and people engage in an increasingly digital world.
Aug. 29, 2015 09:30 AM EDT Reads: 430
Introducing Containers & Microservices Bootcamp at @CloudExpo Silicon Valley | #Containers #Microservices
SYS-CON Events announced today the Containers & Microservices Bootcamp, being held November 3-4, 2015, in conjunction with 17th Cloud Expo, @ThingsExpo, and @DevOpsSummit at the Santa Clara Convention Center in Santa Clara, CA. This is your chance to get started with the latest technology in the industry. Combined with real-world scenarios and use cases, the Containers and Microservices Bootcamp, led by Janakiram MSV, a Microsoft Regional Director, will include presentations as well as hands-on...
Aug. 29, 2015 09:15 AM EDT Reads: 161
The pricing of tools or licenses for log aggregation can have a significant effect on organizational culture and the collaboration between Dev and Ops teams. Modern tools for log aggregation (of which Logentries is one example) can be hugely enabling for DevOps approaches to building and operating business-critical software systems. However, the pricing of an aggregated logging solution can affect the adoption of modern logging techniques, as well as organizational capabilities and cross-team ...
Aug. 29, 2015 08:30 AM EDT Reads: 360
In today's digital world, change is the one constant. Disruptive innovations like cloud, mobility, social media, and the Internet of Things have reshaped the market and set new standards in customer expectations. To remain competitive, businesses must tap the potential of emerging technologies and markets through the rapid release of new products and services. However, the rigid and siloed structures of traditional IT platforms and processes are slowing them down – resulting in lengthy delivery ...
Aug. 29, 2015 07:45 AM EDT Reads: 551
Skeuomorphism usually means retaining existing design cues in something new that doesn’t actually need them. However, the concept of skeuomorphism can be thought of as relating more broadly to applying existing patterns to new technologies that, in fact, cry out for new approaches. In his session at DevOps Summit, Gordon Haff, Senior Cloud Strategy Marketing and Evangelism Manager at Red Hat, discussed why containers should be paired with new architectural practices such as microservices rathe...
Aug. 29, 2015 06:00 AM EDT Reads: 371
Any Ops team trying to support a company in today’s cloud-connected world knows that a new way of thinking is required – one just as dramatic than the shift from Ops to DevOps. The diversity of modern operations requires teams to focus their impact on breadth vs. depth. In his session at DevOps Summit, Adam Serediuk, Director of Operations at xMatters, Inc., will discuss the strategic requirements of evolving from Ops to DevOps, and why modern Operations has begun leveraging the “NoOps” approa...
Aug. 29, 2015 05:30 AM EDT Reads: 342
Puppet Labs has announced the next major update to its flagship product: Puppet Enterprise 2015.2. This release includes new features providing DevOps teams with clarity, simplicity and additional management capabilities, including an all-new user interface, an interactive graph for visualizing infrastructure code, a new unified agent and broader infrastructure support.
Aug. 29, 2015 03:00 AM EDT Reads: 474
DevOps has traditionally played important roles in development and IT operations, but the practice is quickly becoming core to other business functions such as customer success, business intelligence, and marketing analytics. Modern marketers today are driven by data and rely on many different analytics tools. They need DevOps engineers in general and server log data specifically to do their jobs well. Here’s why: Server log files contain the only data that is completely full and accurate in th...
Aug. 29, 2015 12:15 AM EDT Reads: 332
It’s been proven time and time again that in tech, diversity drives greater innovation, better team productivity and greater profits and market share. So what can we do in our DevOps teams to embrace diversity and help transform the culture of development and operations into a true “DevOps” team? In her session at DevOps Summit, Stefana Muller, Director, Product Management – Continuous Delivery at CA Technologies, answered that question citing examples, showing how to create opportunities for ...
Aug. 28, 2015 12:00 PM EDT Reads: 447
Several years ago, I was a developer in a travel reservation aggregator. Our mission was to pull flight and hotel data from a bunch of cryptic reservation platforms, and provide it to other companies via an API library - for a fee. That was before companies like Expedia standardized such things. We started with simple methods like getFlightLeg() or addPassengerName(), each performing a small, well-understood function. But our customers wanted bigger, more encompassing services that would "do ...
Aug. 28, 2015 11:00 AM EDT Reads: 177
Culture is the most important ingredient of DevOps. The challenge for most organizations is defining and communicating a vision of beneficial DevOps culture for their organizations, and then facilitating the changes needed to achieve that. Often this comes down to an ability to provide true leadership. As a CIO, are your direct reports IT managers or are they IT leaders? The hard truth is that many IT managers have risen through the ranks based on their technical skills, not their leadership ab...
Aug. 28, 2015 10:00 AM EDT Reads: 277
Whether you like it or not, DevOps is on track for a remarkable alliance with security. The SEC didn’t approve the merger. And your boss hasn’t heard anything about it. Yet, this unruly triumvirate will soon dominate and deliver DevSecOps faster, cheaper, better, and on an unprecedented scale. In his session at DevOps Summit, Frank Bunger, VP of Customer Success at ScriptRock, will discuss how this cathartic moment will propel the DevOps movement from such stuff as dreams are made on to a prac...
Aug. 28, 2015 09:45 AM EDT Reads: 184
SYS-CON Events announced today that DataClear Inc. will exhibit at the 17th International Cloud Expo®, which will take place on November 3–5, 2015, at the Santa Clara Convention Center in Santa Clara, CA. The DataClear ‘BlackBox’ is the only solution that moves your PC, browsing and data out of the United States and away from prying (and spying) eyes. Its solution automatically builds you a clean, on-demand, virus free, new virtual cloud based PC outside of the United States, and wipes it clean...
Aug. 28, 2015 09:45 AM EDT Reads: 341
In his session at 17th Cloud Expo, Ernest Mueller, Product Manager at Idera, will explain the best practices and lessons learned for tracking and optimizing costs while delivering a cloud-hosted service. He will describe a DevOps approach where the applications and systems work together to track usage, model costs in a granular fashion, and make smart decisions at runtime to minimize costs. The trickier parts covered include triggering off the right metrics; balancing resilience and redundancy ...
Aug. 28, 2015 09:30 AM EDT Reads: 129
The Microservices architectural pattern promises increased DevOps agility and can help enable continuous delivery of software. This session is for developers who are transforming existing applications to cloud-native applications, or creating new microservices style applications. In his session at DevOps Summit, Jim Bugwadia, CEO of Nirmata, will introduce best practices, patterns, challenges, and solutions for the development and operations of microservices style applications. He will discuss ...
Aug. 27, 2015 02:15 PM EDT Reads: 506