- 1、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。。
- 2、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。
- 3、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
* * Emphasize the benefits from reservations – recycling of machines, never losing track, fewer machines needed, low hanging fruit way to share. Emphasize that we do the back-end capacity reservation “look-ahead” to guarantee resources to maximize capacity usage On-demand – typically used for cycle stealing or opportunistic computing, or overflow to EC2 or external hoster. * I can provide a detail summary here, or walk thru the following slides. * * * * Node with green colors means free * Node with green colors means free * * We have something call “PTM” which allows specification of “if then else” rules and these rules can be used to optimize for energy consumptions. * * * * * We use a wizard approach to help the admin to define service offerings quickly * And very easy for the “end-user” or customer to request a predefined set of offerings. Just fill in, and you get the application environments in the next slide … * Some out-of-the-box reports (I emphasize some, think of these reports as basic ones), but all data in the system is also logged into a SQL database which allows the IT to do their own chargeback and analysis which is the more typical usage of our system. * * Challenge: Expanding compute capacity while reducing costs Dedicated SMP hardware – expensive, not scalable, provisioned for peak Seven Major Trading and Risk Systems – expensive to maintain, difficult to scale Solution: Built a grid infrastructure combining 7 major trading and risk systems, now expanded to 21 apps. 1 million dollars saved in first year (2003) millions more over the subsequent years. “The grid approach allows the bank to reduce operational risk, for instance when an isolated server fails, you have an ability to respond much more flexibly. Michael AshworthCIO, JPMC * * * * Topology Aware (基于网络拓扑)的分配策略 机架二 机架一 刀片中心一 刀片中心二 刀片中心三 刀片中心四 可用的节点 被占用的节点 尽量保证为应用程序分配的服务器资源之间的网络路径最小 仅在万不得已的情况下才跨刀片中心、机架、交换机分配服务器资源 提高应用程序运行效率 * Topology Aware (基于网络拓扑)的分配策略 机架二 机架一
原创力文档


文档评论(0)