Editor’s Note: P99 CONF 2022 is now a wrap! You can (re)watch all videos and access the decks now.
ACCESS ALL THE VIDEOS AND DECKS NOW
P99 CONF is now days away. We’ve already shared a bit about what you can expect, and we’re busy setting the virtual stage for 50+ expert engineers to share innovations, insights, and best practices for lowering latency and optimizing performance from a broad range of perspectives– programming languages, CPUs, front-ends, back-ends, and observability.
To peak pique your interest, let’s hear directly from some of the speakers you can learn from and interact with live on October 19 and 20. What are they most excited to share – and learn?
Instant Access (Available Anytime, Both Days)
Roman Gershman, DragonflyDB
Talk: How Dashtable Helps Dragonfly Maintain Low Latency
Why attend your session – in 7 words or less?
Learn the art of optimizing hashtable latency.
What other session on the agenda are you most looking forward to and why?
I am interested to listen to “Optimizing Servers for High-Throughput and Low-Latency” by Alexey Ivanov. As backend developers, we strive to build the fastest software possible, but we often neglect the low-hanging fruit – OS configuration. I am always keen to learn up-to-date techniques for fine-tuning the networking stack.
Extra credit: Any recommended resources for getting more out of your session?
If you have not heard about the dragonfly project, please visit https://github.com/dragonflydb/dragonfly and read our story. Also, I warmly suggest reading our introductory article about DashTable at https://github.com/dragonflydb/dragonfly/blob/main/docs/dashtable.md. It
gives a decent overview of DashTable, and why it is more memory efficient than other approaches.
Yuying Song, PingCAP
Talk: How We Reduced Performance Tuning Time by Orders of Magnitude with Database Observability
Why attend your session – in 7 words or less?
Tuning database by DB time and color.
What other session on the agenda are you most looking forward to and why?
Gil Tene’s Misery Metrics & Consequences because I’m curious what “misery metrics” means, how to measure them, and how to use them for performance tuning.
Extra credit: Any recommended resources for getting more out of your session?
1. The two performance tuning methodologies mentioned in my session are “tuning by database time” and “tuning by color”. Our team has described them in detail in TiDB Performance Tuning Guide (https://docs.pingcap.com/tidb/dev/performance-tuning-methods). Please refer to the tuning guide for more details.
2. The TiPerformance Channel on YouTube is a collection of short sessions presented by my teammates and me to share with you the performance tips we learned from real-world workloads. We have just released the first session: Find Performance Problems using the TiDB Performance Dashboard (https://www.youtube.com/watch?v=-hIkQE8YL-4), which is related to my session on P99 CONF. The second session will be released in October. Stay tuned!
Oren Eini, RavenDB
Talk: C# as a System Language
Why attend your session – in 7 words or less?
Learn how C# can be used for high perf systems.
What other session on the agenda are you most looking forward to and why?
Square Engineering’s “Fail Fast, Retry Soon” Performance Optimization Technique. Failure handling is easy, it’s obvious when something is wrong. Latency reduction is a lot harder, since there is no single culprit. I’m very interested in anything that can help in this regard.
Extra credit: Any recommended resources for getting more out of your session?
I’m talking a lot about how to apply those sort of details in this video about the internals of RavenDB: https://youtu.be/8mCsxcvXjp4
Michael Barker, Ephemeris Consulting
Talk: Fast and Fault Tolerant
Why attend your session – in 7 words or less?
Build services with stellar performance and reliability.
What other session on the agenda are you most looking forward to and why?
Sharpening the Axe: The Primacy of Toolmaking. I’m always fascinated by what Oxide Computer are working on and I would like to get their insight on the balance of building tool vs moving the product forward.
Extra credit: Any recommended resources for getting more out of your session?
Check out http://github.com/real-logic/aeron for an open source framework for building services as replicated state machines. Also look at http://weareadaptive.com for commercial support, consulting, and additional tooling around Aeron.
Roni Dover, Digma
Talk: The Observant Developer — OpenTelemetry from Code to Prod
Why attend your session – in 7 words or less?
Observability is changing the way we write code.
What other session on the agenda are you most looking forward to and why?
Gil Tene’s keynote on metrics – his “from the trenches” knowledge of how to really measure performance is priceless and practical.
Extra credit: Any recommended resources for getting more out of your session?
Continuous Feedback is a new field that is worth catching up on, we created a slack group at https://continuousfeedback.org/slack where you can find a lot of resources. Of course, you’re also welcome to check out our CF platform at digma.ai!
Day 1 (Wednesday)
Alex Gallego, Redpanda
Talk: Removing Implicit Deadlocks on a Thread-Per-Core Architecture with 2-Phase Processing
Why attend your session – in 7 words or less?
Fix implicit deadlocks in async code.
What other session on the agenda are you most looking forward to and why?
Anything Gil Tene says is worth listening to when it comes to measuring and benchmarking.
Extra credit: Any recommended resources for getting more out of your session?
Checking out our github https://github.com/redpanda-data/redpanda and joining our slack redpanda.com/slack for questions on the session.
Omar Elgabry, Square
Talk: Square Engineering’s “Fail Fast, Retry Soon” Performance Optimization Technique
Why attend your session – in 7 words or less?
To go beyond performance optimization techniques’ happy path.
What other session on the agenda are you most looking forward to and why?
Overcoming Variable Payloads to Optimize for Performance by Armin Ronacher. I am always curious to learn about real use cases to “Optimize for Performance”.
Extra credit: Any recommended resources for getting more out of your session?
My medium blog: https://medium.com/@OmarElgabry
Otherwise, feel free to reach out on LinkedIn for any questions! https://www.linkedin.com/in/omarelgabry/
Peter Zaitsev, Percona
Talk: A Deep Dive into Query Performance
Why attend your session – in 7 words or less?
Query Response Time Matters!
What other session on the agenda are you most looking forward to and why?
“Square Engineering’s “Fail Fast, Retry Soon” Performance Optimization Technique.” It is always great to learn what approaches actually work at scale.
Extra credit: Any recommended resources for getting more out of your session?
Read up on Red Method. Check out the blog post on how you can apply it to databases with a custom dashboard in Percona Monitoring and Management. For an even more fundamental education, make sure to read “The Goal” – one of the most useful books on Database Performance optimization, which does not have a single word about database performance.
Daniele Salvatore Albano, Microsoft
Talk: cachegrand: A Take on High Performance Caching
Why attend your session – in 7 words or less?
High performances and low latencies? Yes, please!
What other session on the agenda are you most looking forward to and why?
Removing Implicit Deadlocks on a Thread-Per-Core Architecture with 2-Phase Processing – This is an approach that has been explored in cachegrand as in some rare conditions, depending on the user application, a user-level deadlock might happen and although cachegrand has an automatic abort system to deal with these cases it would impact the performances, implementing a 2 step processing algorithm to avoid it entirely is a huge win-win.
Using eBPF for High-Performance Networking in Cilium – I use eBPF to debug and analyze the performances in cachegrand and it’s always interesting to see how other people use it, where they place hooks, which metrics they collect, etc.
Why User-Mode Threads Are Good for Performance – Fibers, or User-Mode threads, are a key component in cachegrand so seeing other usage of this approach is great. On top of it, I love the fact that Oracle is “resurrecting” the User-Mode threads in the JVM as I think they are really the key to provide amazing performances without using the “callback hell” approach where you have to pass nested callbacks of callbacks everywhere together with a nested status of everything.
Extra credit: Any recommended resources for getting more out of your session?
Caching and Cache-Efficient Algorithms from MIT ( https://www.youtube.com/watch?v=xDKnMXtZKq8 )
Scott Meyers – Cpu Caches and Why You Care (https://www.youtube.com/watch?v=WDIkqP4JbkE)
Faster: A Concurrent Key-Value Store with In-Place Updates (https://www.microsoft.com/en-us/research/uploads/prod/2018/03/faster-sigmod18.pdf)
https://graphitemaster.github.io/fibers/ – It covers some concepts explained in cachegrand (the fibers) from “zero-to-hero” and it’s also relevant for the Oracle talk on the user-threads I referenced above.
Alex Hidalgo, Nobl9
Talk: Throw Away Your Nines
Why attend your session – in 7 words or less?
You Might Be Doing It Wrong (Maybe).
What other session on the agenda are you most looking forward to and why?
Omar Elgabry’s “Square Engineering’s “Fail Fast, Retry Soon” Performance Optimization Technique.” We often spend too much time focused on making things more reliable on the first attempt when really we can often make things more reliable by failing fast and retrying.
Extra credit: Any recommended resources for getting more out of your session?
The SLO Book (https://amazon.com/Implementing-Service-Level-Objectives-Practical/dp/1492076813)
Jeff Utter, theScore
Talk: Aggregator Leaf Tailer: Bringing Data to Your Users with Ultra Low Latency
Why attend your session – in 7 words or less?
Materialize Kafka with Aggregator Leaf Tailer. Fast!
What other session on the agenda are you most looking forward to and why?
Gil Tene’s Keynote: Misery Metrics & Consequences – Gil’s work with wrk2 and coordinated omission changed the way I view latency metrics! I’m looking forward to what he has to share.
Filipe Oliveira: End-To-End Performance Testing, Profiling, and Analysis at Redis – Automated end to end performance testing is difficult and something every company I have worked at has dreamed about but never quite achieved. I’m excited to hear how Redis Labs is doing it!
Extra credit: Any recommended resources for getting more out of your session?
Learn about the things that Rockset is doing with Aggregator Leaf Tailer, such as https://rockset.com/blog/aggregator-leaf-tailer-an-architecture-for-live-analytics-on-event-streams/
This isn’t exactly what we’ve built but a good general primer on the topic.
Day 2 (Thursday)
Leandro Melendez (Senor Performo), Grafana k6
Talk: Continuous Performance from Load Testing to SRE and Beyond
Why attend your session – in 7 words or less?
Understand why your old performance is failing.
What other session on the agenda are you most looking forward to and why?
Omar Elgabry’s – Omar Elgabry’s – “Fail Fast, Retry Soon” Performance Optimization – I like the perspective of trial and error on performance issues. In the end, most of the time that is what we do to detect, diagnose and re-test that performance is right. The description of a real life situation makes this even more interesting.
Extra credit: Any recommended resources for getting more out of your session?
Book: The hitchiking guide to load testing projects: https://amzn.to/37wqpyx
YouTube: https://www.youtube.com/c/se%C3%B1orperformoeng
K6-Learn: github.com/grafana/k6-learn
Henrik Rexed, Dynatrace
Talk: Why Kubernetes Freedom Requires Chaos Engineering to Shine in Production
Why attend your session – in 7 words or less?
Avoid production outage due to K8s settings! Sure I’m in.
I know it’s more than 7 words 😉
What other session on the agenda are you most looking forward to and why?
The are several amazing speakers that cover really interesting sessions:
Using eBPF for High-Performance Networking in Cilium from Liz Rice. I always been a big fan of the ebpf technology , and I think Cilium was able to build an amazing solution to secure our network without adding complexity
Performance Tuning with Precision from Charity Majors
Measuring P99 Latency in Event-Driven Architectures with OpenTelemetry , Anton Rodrigues. Tracing an event driven architecture has always been a challenge, so I’m hoping to learn out of this session
Extra credit: Any recommended resources for getting more out of your session?
My presentation is summary of 2 distinct episode of IsitObservable. So, this session will explain the value of using Chaos engineering in K8S. If you are looking on deeper content on this topic you should watch https://youtu.be/mwu5eLgUKq4 and https://youtu.be/qLIF8-xGrqc.
Piotr Sarna, ScyllaDB
Talk: Keeping Latency Low for User-Defined Functions with WebAssembly
Why attend your session – in 7 words or less?
WebAssembly is the future.
What other session on the agenda are you most looking forward to and why?
I’m looking forward to Avi Kivity’s deep dive into how disks behave in the database world. For people who always want to understand how everything works under the hood (like me), I expect it to be a very educational experience
Extra credit: Any recommended resources for getting more out of your session?
https://github.com/scylladb/scylladb/pull/11351
Mark Gritter, Akita Software
Talk: Taming Go’s Memory Usage — and Avoiding a Rust Rewrite
Why attend your session – in 7 words or less?
Memory allocation is unpredictable! How to cope.
What other session on the agenda are you most looking forward to and why?
I’m interested in Bryan Cantrill’s talk on purpose-built tooling. I’ve built and used custom tools, but I’m interested in how to build a culture that encourages tool-building.
Sagy Volkov, Lightbits Labs
Talk: Testing Persistent Storage Performance in Kubernetes with Sherlock
Why attend your session – in 7 words or less?
Performance is the key to infrastructure stability.
What other session on the agenda are you most looking forward to and why?
A deep dive into query performance. Peter Zaitsev. I have a lot of history being a production DBA and managing DBAs before shifting into the storage world and I always find the art of tuning a query very fascinating.
Extra credit: Any recommended resources for getting more out of your session?
Don’t believe me or any performance paper, go and try it yourself 🙂
Felipe Huici, Unikraft
Talk: It’s Time to Debloat the Cloud with Unikraft
Why attend your session – in 7 words or less?
Cloud costs are a royal pain.
What other session on the agenda are you most looking forward to and why?
“Analyze Virtual Machine Overhead Compared to Bare Metal with Tracing”. It’s generally accepted that virtual machines add overhead compared to running bare metal, but how much and where does it come from exactly? I’m really looking forward to this talk to provide precise answers to these questions.
Extra credit: Any recommended resources for getting more out of your session?
Please join our open source community by visiting www.unikraft.org and/or meet us on our Discord server at https://bit.ly/UnikraftDiscord. For commercial solutions please check out our company website at https://unikraft.io/.
Alexey Ivanov, Dapper Labs
Talk: Optimizing Servers for High-Throughput and Low-Latency
Why attend your session – in 7 words or less?
Interested in low-level performance tuning but were afraid to ask?
(I fail at counting to seven=))
What other session on the agenda are you most looking forward to and why?
Keynotes!! and “Taming Go’s Memory Usage — and Avoiding a Rust Rewrite.”
Extra credit: Any recommended resources for getting more out of your session?
https://twitter.com/SaveTheRbtz
Brian Likosar, StormForge
Talk: The Role of Machine Learning In Cloud Native Performance Optimization
Why attend your session – in 7 words or less?
One real way ML helps Kubernetes perform!
What other session on the agenda are you most looking forward to and why?
I’m looking forward to is the keynote on Performance Tuning with Precision by Honeycomb. I’m a fan of anything Charity Majors is involved in and I can’t wait to see how she and Ian cover the topic. I’m sure I’ll learn something new.
Extra credit: Any recommended resources for getting more out of your session?
Watch this video which is a nice overview of what we do here at StormForge
https://youtu.be/LZSxe4RRZaE