Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.

Dmitry Tiagulskyi, Yaroslav Yermilov "It Scales Until It Doesn’t"

335 vues

Publié le

We are used to thinking that “high-load” means distributed systems, computing power, application, and kernel profiling. But sometimes you can’t simply scale your cluster. Maybe your hashmaps don’t fit in the server memory. Maybe you need single-digit millisecond latency. Maybe the cost is too high. Or your server is a … mobile phone.

In this talk, we will show how popular and lesser-known algorithms, data structures, and systems tuning helped us to overcome these blockers. Who said you don’t need to know algorithms nowadays?

Publié dans : Technologie
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Dmitry Tiagulskyi, Yaroslav Yermilov "It Scales Until It Doesn’t"

  1. 1. It Scales Until It Doesn’t yaroslav.yermilov@grammarly.com - Software Engineer @ Core Services dima.tiagulskyi@grammarly.com - Software Engineer @ Core Services
  2. 2. ● Algorithms & Data Structures for Language Models (Ngrams) ○ Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, Clifford Stein “Introduction to Algorithms” (Chapter 11 “Hash Tables” (253) specifically) ○ F.C. Botelho, R. Pagh, and N. Ziviani. Simple and space-efficient minimal perfect hash functions. In Proc. of the 10th Workshop on Algorithms and Data Structures (WADs’07), pages 139–150. Springer LNCS vol. 4619, 2007. ○ Djamal Belazzougui, Fabiano Botelho, and Martin Dietzfelbinger. 2009. Hash, displace, and compress. Algorithms - ESA 2009, pages 682–693. ○ Kenneth Heafield. 2011. KenLM: Faster and smaller language model queries. In Proceedings of the Sixth Workshop on Statistical Machine Translation, Edinburgh, UK, July. Association for Computational Linguistics. ○ Adam Pauls and Dan Klein. 2011. Faster and smaller ngram language models. In Proceedings of ACL, Portland, Oregon. ○ David Talbot and Miles Osborne. 2007. Randomised language modelling for statistical machine translation. In Proceedings of ACL, pages 512–519, Prague, Czech Republic. ○ D. Guthrie, M. Hepple, and W. Liu, “Efficient minimal perfect hash language models,” in Proceedings of LREC’10. Valletta, Malta: European Language Resources Association (ELRA), May 2010. References
  3. 3. ● AWS Virtualization, ENA ○ https://en.wikipedia.org/wiki/X86_virtualization ○ https://aws.amazon.com/ru/blogs/aws/elastic-network-adapter-high-performance-network-interface-for-amazon-ec2/ ○ https://medium.com/@paccattam/aws-enhanced-networking-an-overview-aee8a852cf5c ○ AWS re:Invent 2017: Optimizing Network Performance for Amazon EC2 Instances: https://youtu.be/-dWgqtGKPfc ○ https://www.kernel.org/doc/Documentation/networking/scaling.txt ○ AWS re:Invent 2017: C5 Instances and the Evolution of Amazon EC2 Virtualization: https://youtu.be/LabltEXk0VQ ○ http://www.brendangregg.com/blog/2017-11-29/aws-ec2-virtualization-2017.html ○ https://blog.ubuntu.com/2017/04/05/ubuntu-on-aws-gets-serious-performance-boost-with-aws-tuned-kernel References
  4. 4. ● NUMA ○ https://www.redhat.com/files/summit/2014/summit2014_riel_chegu_w_0340_automatic_numa_balancing.pdf ○ https://www.cmg.org/wp-content/uploads/2015/10/numa.pdf ● Network Servers, IO Multiplexing & Epoll ○ https://eli.thegreenplace.net/2018/measuring-context-switching-and-memory-overheads-for-linux-threads/ ○ Asynchronous IO with Boost.Asio: https://youtu.be/rwOv_tw2eA4 ○ https://eklitzke.org/blocking-io-nonblocking-io-and-epoll ○ https://habr.com/post/416669/ ● Netty ○ One Framework to rule them all by Norman Maurer: https://youtu.be/DKJ0w30M0vg ○ https://dou.ua/lenta/columns/netty-optimization/ References
  5. 5. ● Benchmarking ○ http://www.brendangregg.com/blog/2018-06-30/benchmarking-checklist.html ○ http://www.brendangregg.com/activebenchmarking.html ○ http://www.brendangregg.com/usemethod.html ○ http://www.brendangregg.com/linuxperf.html ● Java Profiling ○ http://psy-lob-saw.blogspot.com/2016/02/why-most-sampling-java-profilers-are.html ○ https://github.com/jvm-profiling-tools/async-profiler ○ https://medium.com/netflix-techblog/java-in-flames-e763b3d32166 References