如何设计高可靠性的分布式并行系统.pptx

  1. 1、本文档共57页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
如何设计高可靠性的分布式并行系统;How can we understand complex software?;The BIG picture;;;;Basic Properties;To program systems of communicating objects we need to make it easy to write parallel programs;?;It’s actually difficult to write parallel programs so …;Make it easy to write concurrent programs;Automate the parellelization of concurrent programs;Programmer decides the;Observe the world and the communication channels;;;;What are the messages?;One parallel operation in real world;;One parallel operation in real world;;How do we make it easy to write concurrent programs?;have a language with only 3 concurrency primitives;spawn;shared memory;Pure message passing;Why Pure Message Passing?;;One programming model;One Programming Model;Obeys the;Which laws of physics?;Details;spreading processes is difficult;OTP team at Ericsson (2-3 people know the multi-core part);100K programmers?? know nothing about multi-core;Their programs should run 0.75 x N times faster on N core computers;Why Pure Message Passing?;Failure;;P(fail) = 10^-3 P(fail)^2 = 10^-6 100 9’s reliability with 34 computers The key is independence No dangling pointers No shared memory No synchronous events;Scalability;Key is independent isolated computations;Shared memory is evil;Does this work?;Yes WhatsApp, Klarna, Ericsson, …;Erlang is also;Values are immutable;Fault tolerance with immutable values;;The BEAM;Inside the Beam;Learning;;

文档评论(0)

我是大神 + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档