AWS re:Invent 2017: How Netflix Tunes Amazon EC2 Instances for Performance (CMP325)
Amazon Web ServicesCited 3× on HN, 2018–2022
talk · advanced · 2018–2022 on HN
A Netflix performance engineer presents how Netflix tunes EC2 instances, covering instance type choice, Xen modes, SR-IOV and kernel tuning. The goal is near bare-metal performance in the cloud.
HN cites it as an example of workload-dependent performance tuning at scale and of Netflix's AWS practices.
Up next
10 mo on HNVelocity NY 2013: Richard Cook, "Resilience In Complex Adaptive Systems"
O'Reilly
4 HN recommender · 2017–2021
11 mo on HNDebugging Under Fire: Keep your Head when Systems have Lost their Mind • Bryan Cantrill • GOTO 2017
GOTO Conferences
8 HN recommender · 2018–2025
9 mo on HNAWS re:Invent 2018: Amazon DynamoDB Deep Dive: Advanced Design Patterns for DynamoDB (DAT401)
Amazon Web Services
12 HN recommender · 2019–2025
9 mo on HNLiterate Devops with Emacs
Howard Abrams
8 HN recommender · 2015–2025
6 mo on HNContainers From Scratch • Liz Rice • GOTO 2018
GOTO Conferences
6 HN recommender · 2018–2025
6 mo on HNOver 200,000 Servers in One Place! Visiting Hetzner in Falkenstein (Germany)
der8auer EN
6 HN recommender · 2021–2025
5 mo on HNBryan Cantrill from Joyent on Manta: internet-facing object storage facility that features compute
AI Council
4 HN recommender · 2014–2026
6 mo on HNZebras All the Way Down - Bryan Cantrill, Uptime 2017
Code & Supply
6 HN recommender · 2018–2025
5 mo on HNElixirConf 2018 - Docker and OTP Friends or Foes - Daniel Azuma
ElixirConf
4 HN recommender · 2018–2023
6 mo on HNStanford Seminar - The Soul of a New Machine: Rethinking the Computer
Stanford Online
3 HN recommender · 2020–2025
5 mo on HNAWS re:Invent 2016: Tuesday Night Live with James Hamilton
Amazon Web Services
5 HN recommender · 2016–2023
5 mo on HNHow We've Scaled Dropbox
Stanford
5 HN recommender · 2014–2023
3 comments from Hacker News
it was actually 5x, my bad, but here's the references: Original info being mentioned as a rumour (2016): link Slides for confirmation by Greeg (2017): link Talk (2017): ▶ this video More discussion: link Original tweet has been deleted, but I have a screenshot here slide 9: link
Well, most distributions choose saner defaults. For instance, RHEL/CentOS/AmazonLinux use noop io-scheduler by default for SSD based machines. Though, I get your point: for instance, the TCP keep-alive defaults aren't suitable for all kinds of workloads: link I guess, that's why performance tuning is an art form as it is workload dependant: ▶ this video
They use a lot of AWS services - not just virtual machines - First they use Ubuntu Linux - SES (Email) - ElasticSearch (AWS has their own managed version) - SQS (queueing system) - S3 (storage) -auto scaling Most of the optimization's they do are standard things at scale where they are tuning by measuring performance. ▶ this video and the slide deck from the presentation. link But that's even more of a reason to choose AWS, Netflix has open sourced dozens of tools specifically related to AWS. You get to take advantage of their tools and knowledge.