Presentation: Scaling Up Performance Benchmarking

Track: Bare Knuckle Performance

Location: Pacific LMNO

Duration: 10:35am - 11:25am

Day of week:

Slides: Download Slides

Level: Advanced

Persona: Backend Developer, Developer

This presentation is now available to view on InfoQ.com

Watch video with transcript

Abstract

Performance scaling along with responsiveness is very important to both on-premise as well as cloud deployments. Often, deployments can be very simple to an extremely complex web of services where there is no single magic bullet for performance scaling. As part of the development of a very complex benchmark, we learned many lessons related to scaling up, scaling out and responsiveness. We would like to share application architecture decisions, observations points, problem analysis methodologies etc. which can be applied when architecting, deploying and analyzing real production applications. We will share how application level, as well as system level metrics, change for various application architecture decisions as well as high-level JVM analysis to correlate tail latencies.

Speaker: Anil Kumar

Datacenter Performance Architect @Intel

Anil is one of the earliest contributors to Java virtual machines GC, large pages, profiling etc. He has extensive experience in performance optimizations as well as benchmarks development and standardization. He represented Intel for 5 years in the JCP Executive Committee (Java Community Process) and currently chairs the SPEC OSGjava committee which owns the benchmarks SPECjms2007, SPECjvm2008, SPECjEnterprise2010, SPECjbb2015 and SPECjEnterprise2018 Web Profile. Anil has been one of the key architects for SPECjbb2005, SPECjvm2008, SPECpower_ssj2008 and SPECjbb2015 benchmarks as well as architect for workloads at https://github.com/Node-DC.

Anil has delivered many talks at conferences like JavaOne and keynotes at SouJava Brazil, ICPE in Dresden, Node Summit 2017 and Node Interactive 2017. Currently he is also leading evaluation of data center scale end-to-end workloads and benchmarks. His next dream is to design and architect benchmarks which can represent latest production deployments and adapt to faster evolving ecosystems including new architectures for AI & ML, ASICs and accelerators. In his spare time, he loves mentoring young kids for STEM and judging science fairs.

Find Anil Kumar at

Speaker: Monica Beckwith

Java Champion, First Lego League Coach, passionate about JVM Performance @Microsoft

Java Champion Monica Beckwith is considered a subject matter expert, has several published articles and gets regular invitations to give talks on JVM/JIT Compilation/Garbage Collection (GC). She is also a JavaOne Rock Star.

Monica has made various performance contributions to the Java HotSpot VM by identifying the need for a NUMA-aware allocator and allocation patterns, reduction of redundant instructions, reduction of the Java object header, prefetching patterns, redundant array checks in a loop and various other optimizations for the JIT compiler, the generated code, the JVM heuristics and garbage collection and collectors.

Prior to joining Microsoft, Monica was the JVM Performance Architect at Arm. Her past also includes leading Oracle’s Garbage First Garbage Collector performance team.

Find Monica Beckwith at

Similar Talks

JIT vs AOT Performance With GraalVM

Qcon

Developer Advocate for GraalVM @Oracle

Alina Yurenko

Parsing JSON Really Quickly: Lessons Learned

Qcon

Professor and Department Chair @TELUQ - Université du Québec

Daniel Lemire

Fault Tolerance at Speed

Qcon

Ex-NASA Researcher and High Performance Distributed Systems Whisperer

Todd Montgomery