The Third Community Over Code Performance Engineering Track, Halifax, 2023
On Sunday 8 Oct 2023 we had another successful Performance Engineering Track at the Community Over Code Conference in Halifax, Canada.
Thanks to the co-chair Roger Abelenda, the speakers and about 120 attendees (20% up from last year) for participating and helping make it a great event! Also thanks to everyone who submitted proposals; as usual we only had room for 6 talks from around 14 submissions (< 50% acceptance rate).
Here are some highlights from the talks.
Talk 1 Paul Brebner Developing Fast Applications With Open Source Software — Without The Fury (Kafka)
This was the 2nd time I’ve given my “everything I know about Apache Kafka application performance” talk (the 1st time was Community Over Code Asia 2023 in Beijing) — a bit of fun with cars + Kafka.
Talk 2 Ritesh H. Shukla, Duong Nguyen, Tanvi Penumudy Design patterns and then the road to realize billions of objects, and exabytes of capacity, while preserving performance in Apache Ozone
The use of simulation to prove the scalability of a system (which couldn’t easily be checked by normal benchmarking due to hardware constraint) was noteworthy — I’ve been on the lookout for the use of simulation for open source software for a while, and I think this is the 1st example I’ve seen — congratulations! Also the use of architectural tradeoffs to improve the scalability of a system was interesting.
Talk 3 German Eichberger, Pallavi Iyengar Performance measurement and tuning of Cassandra 5.0 transactions on Cloud infrastructure
This talk emphasized that there is a need for better benchmarking tools for testing NoSQL transaction performance (E.g. for Cassandra).
Talk 4 Otavio Rodolfo Piske Hunting Performance Monsters on the Back of a Camel
A nice Monster Hunting story and pictures! Also the benefits of using multiple approaches/sources of data to find and confirm the source of performance issues is valuable for other projects. I also think this is the first time I’ve come across using a source code/byte code analysis tool to find potential issues — to find potential type pollution (and of course you need access to the source code to fix them) — both are obvious benefits of performance engineering with open source software. Also the observation that software engineers are not primarily performance engineers so you need multiple tools and sufficient data to ensure changes are actually critical for performance improvement.
Talk 5 Roger Abelenda Quick load testing from Selenium scripts
A very nice demonstration of Selenium and JMeter DSL for load testing, including pictures referencing the movie “Evolution”!
Talk 6 Stefan Vodita Lessons Learned from Benchmarking Amazon’s E-commerce Search Engine
After 25 years in performance engineering I don’t often learn many new things, but this talk had a good introduction to the theory and practice of making sure you document your benchmarking goals, and then use the correct theory and tools to tell you if the candidate code is actually better/worse than the baseline code. Useful ideas included A/A testing to check for the level of noise and bias in just running the benchmarks and the code/platform itself, and A/B testing to check if the results are actually significant.
What was new for me was the use of Bayes’ Theorem instead of, say, a T-Test to determine if two distributions are really different or not. These look novel, and it would be nice to include them in benchmarking tools — maybe something that could be added to JMeter in the future. I think other Apache and Open Source projects could benefit from the best practices and lessons learned from the Apache Lucene community in particular.
The slides are now available from the conference web site (link to slides after each talk abstract). And some photos from the track are here.
The Performance Engineering track is on for Community over Code Europe 2024 too! Here’s the CFP and the conference website/CFP — please consider submitting a talk proposal.
Links to the previous tracks are here:
- Summary of the 1st Apache Con Performance Engineering track from 2022.
2. The 2nd Community Over Code Performance Engineering track (Beijing, 2023).
3. 3rd Community Over Code Performance Engineering track talks and schedule.
So, was there a performance (or at least size) aspect to the harbor photograph? Yes! Did you wonder what the big yellow pipes on the ship (big barge) to the left of the intro photo are for? So did I — the answer = enormous offshore (sea) wind turbine bases! Check out the photo on the CNBC website for an idea of how they are used in practice.
And by way of contrast, the cow painting was by a famous Canadian Folk artist, Maud Lewis — her very tiny house is on display in the local art gallery.
