Performance Testing of NetflixOSS for the Cloud Prize

Where is everyone? I've been writing blogging and tweeting about performance and capacity planning for many years, so I thought I was connected to one or two performance folks....

When we setup the NetflixOSS cloud prize I created a category just for you:

Best Contribution to Performance Improvements

New performance testing tools, results of performance tests and performance oriented bugs. Since this may be aggregated across many pull requests, they should be listed and documented in the Submission.

While there are some interesting contestants in other categories, I've seen very little activity in this one.

There's $10K cash and $5K of AWS credits waiting for someone to do enough work to claim them!

It takes a bit of resource to run tests, so I also have a $100 AWS gift card that I will pass on to the first person to make an entry in this category, regardless of whether they win the prize or have generated any useful data at that point. Make a valid submission to the Cloud Prize by forking the github repo, editing the Submission.md file to describe your entry, entering your details (email/address etc) in the Mailchimp form, and then post a comment here claiming the $100 AWS card, and I'll mail it to you.

After the first one, if you make a case for having generated lots of good data, I'll get some more $100 AWS cards.

Here's some performance project ideas.

• Run throughput and latency stress tests on the Zuul API gateway on various instance types
• Run stress tests for sizing data against other service level projects - Eureka, Edda, Ice, etc.
• Make the cass_jmeter project better in some way
• Analyze the code paths in Astyanax and find ways to make it run faster and use less memory

Go for it, only a few days left. Contest ends at midnight Pacific time Sunday Sept 15th.

Best wishes

Adrian

Book: Release It! Design and Deploy Production Ready Software by Michael Nygard

Essential reading for anyone architecting web sites. It contains clear explanations and real-life anecdotes of many of the patterns and anti-patterns that I've run into over the years, and I'm learning some new ones. Now I just have to get all the engineers on my team to read it... Here it is on Amazon.

Velocity conference summary

Steve Souders writes a good summary of the Velocity Conference. I'd like to present something next year, I have a few ideas already, and have been inspired to try some latency testing projects at work.

Usenix 09 - San Diego in June

I'm presenting again next month, at Usenix 09 in San Diego. I'm giving an all day workshop/tutorial on Tuesday June 16th. I have just updated my slides, combining the talks I have been developing over the years into Solaris/Linux Performance Tools and Tuning, which includes the Capacity Planning with Free Tools material that Mario Jauvin and I developed for CMG.



I have also posted the slides to http://www.slideshare.net/adrianco
.

Please come along, I look forward to seeing you in San Diego.

Cheers Adrian

Speaking at UK CMG TEC Conference - May 19-21st 2008

UKCMG TEC 2008 is near Northampton in the UK. I'm giving two of the same presentations as I gave at the US CMG, the enterprise version of my Millicomputing talk, and a half day workshop on Unix/Linux Performance.

My old friend Phil Harman is also giving a talk on DTrace, and I'm staying in the UK for a few days after the conference to catch up with friends and family.

Patent on Performance and Capacity Planning

Here is another patent that I filed while working at Sun.

System and method for generating a probability distribution of computer performance ratios

A system and method for generating a probability distribution indicating the probability of a range of performance ratios between computer systems. Benchmark and/or other testing results are obtained for both computer systems. For each test, a ratio of one system's performance to the other is calculated. The ratios and/or testing results may be weighted. From the performance ratios, a histogram is produced to indicate the probabilities of achieving the various ratios. Based on the distribution of probabilities, a particular performance ratio can be predicted with a corresponding level of confidence, and used for capacity planning or configuring a computing environment.

Parallel Application Performance and Tuning with OpenMP

I just had dinner with my old friend Ruud van der Pas, who works for Sun's developer tools organization. He is a world class expert in making parallelized applications scale, and recently gave a talk with a very amusing performance conversation.

Ruud is an advocate and expert in OpenMP, and has just published an excellent book on the subject. "Using OpenMP". Published by MIT Press. The work was done together with Barbara Chapman and Gabriele Jost. The ISBN numbers are ISBN-10: 0-262-53302-2 and ISBN-13: 978-0-262-53302-7.

Even if you aren't using OpenMP, the concepts for performance and scaling are the same for everyone...

Computer Measurement Group - CMG07 - San Diego - December 2-7

As usual I will be presenting at CMG07. I have attended every year since 1994, and its the place where I keep track of new developments in the performance management tools industry, as well as learning and sharing with my peers.

CMG moves around the US, last year it was in Reno NV, before that in Orlando FL.

On Sunday 2nd December there is a workshop day. I'm co-presenting (with Mario Jauvin) a half day tutorial on Capacity Planning and Performance Monitoring with Free Tools (for the third year). We updated our slides a bit more this year to track the latest versions of the free tools.

During the week I'm giving a new paper on Millicomputing, that introduces the concepts I have been discussing on my Millicomputing blog, and talks about some of the management challenges of managing large numbers of small computers.

I'm also giving a half day training class on Unix/Linux Performance Management, which is based on the materials I developed over many years in my books and talks. It is still somewhat biased towards Solaris since I know that better, but I have been adding more on Linux and AIX as that is the world I currently live in at work (Netflix runs RedHat Linux and AIX on IBM P-series) and in my Millicomputing sideline (I run Ubuntu on my development host and linux 2.6 on my gumstix).

The conference contains a wide range of content, there is:

• a large vendor exhibition showing the latest tools, vendor presentations and user group meetings to talk about their next steps.
  
• everything from beginner level introductions to deep mathematical theory.
• coverage of networking, storage, CPUs, application level performance.
  
• discussion of Linux, Solaris, AIX, Windows, Mainframe and all combinations
• panel discussions on virtualization, service oriented architectures, scaling large sites

I hope to see some of my readers there. I used to give training classes for Sun on a regular basis, and this is now the main venue for me to present updates versions of those classes.

SEtoolkit and XEtoolkit releases

The SEtoolkit was developed in 1993 by Rich Pettit, and I used it as a way to prototype many new tools and ideas over the years. Its a Solaris specific performance tool scripting language that supports very rapid development of new tools. The SEtoolkit has been widely deployed as the Solaris collector for the popular system performance monitor Orca. Rich gave up development of the SEtoolkit a few years ago, put the code into open source under GPL, and its now available via sourceforge, where it is being maintained by Dagobert Michelsen. A bug in the SEtoolkit was causing it to crash when used with complex disk subsystems, and this has now been fixed in the SE3.4.1 release (April 10th, 2007).

Meanwhile, Rich has been trying to make a multi-platform (Solaris, Windows, Linux, FreeBSD, OSX, HP-UX, AIX) version of SE for a long time, and finally gave up trying to implement his own language, and based his latest development, the XEtoolkit, on Java 5. The first full release XEtoolkit 1.0 came out on April 15th, 2007. The code is released and supported under both open source and commercial licenses, by Rich's new company - Captive Metrics. The GPL license allows full free use of the provided tools, and development of new and derived tools that are also contributed to the community. The commercial license allows custom XEtoolkit development for proprietary tools, with a higher level of support.

The XEtoolkit 1.0 release doesn't support HP-UX or AIX, but AIX support is coming soon. I encourage you to try it out, give Rich some feedback and make it worth his while to continue. He's one of the very best programmers and performance tool architects I've ever met....