浏览 8444 次
精华帖 (0) :: 良好帖 (0) :: 新手帖 (0) :: 隐藏帖 (0)
|
|
---|---|
作者 | 正文 |
发表时间:2007-12-27
注册用户:2000万年增500万 数据库记录:日增量100万条 数据库存储:日增量10G Linux+apache+tomcat/jboss/resin+MySQL 大部分功能是进行交互操作,数据库访问较多 对于此类型网站及数据库该如何设计 CDN Load Balancer(F5) Apache做服务器集群 MySQL做集群 另外数据库表如何设计 例用户表是一张表还是多张表,相应地以活动为例,如果平均每个用户参与了10个活动,这种表的记录数为注册用户数*n条 希望能跟大家在这里互相探讨一下 类似的淘宝,QQ等大型网站的数据库如何设计? 声明:ITeye文章版权属于作者,受法律保护。没有作者书面许可不得转载。
推荐链接
|
|
返回顶楼 | |
发表时间:2007-12-29
这类问题最好直接请人设计了
|
|
返回顶楼 | |
发表时间:2008-01-03
分而治之。把应用分组,每一组服务一定数量的用户,提供一部分数据。
|
|
返回顶楼 | |
发表时间:2008-01-31
大量使用memcache等开源软件+好的数据库设计+集群部署
应该没问题。 |
|
返回顶楼 | |
发表时间:2008-07-21
数据库散列,表散列
个个模块分散到不同的机器上。 |
|
返回顶楼 | |
发表时间:2008-08-14
注册用户:2000万年增500万
......汗一下 |
|
返回顶楼 | |
发表时间:2008-08-19
值得期待大牛们的方案。。。
数据库是不是也要集群啊?好像类似什么数据仓库之类的东东? |
|
返回顶楼 | |
发表时间:2008-08-26
借用另一自贴:
1. If on low budget, use iptables automation to replace F5 2. Avoid using scripting language (Python, PHP, Rails, etc.). Use Java or C if you can afford good programmers; write your own framework if you really have good people that you can keep. 2. Segregate users into different sub-systems and sub-databases - Unless use Oracle RAC (expensive and ACID challenge) like DB clustering, handling your own clusters is a design challenge 3. Read how Yahoo did it http://developer.yahoo.com/performance/rules.html to further optimize the Web site 4. Start think about how to purge old data from day one. (I warned you!) 5. Build in system manageability and monitor-ability from day one. (I warned you!) 6. Optimize database for optimized queries - Export data from the transactional database to a denormalized read-only database that has indexes designed for queries - Decentralize database based on different function - registration/signon vs. transaction 7. Cache data in the memory (before there was DB, all data was stored in memory) - This only works for small amount of data, but could be the most dynamic parts - Write a thread to periodically load data from DB to memory; have Web page use the in memory variables without running SQL; - or have other code update in-memory variables, then write a thread to write data into DB to persist; need to deal with failover scenarios - Not good for CGI or clustered systems 8. Use Web page cache products, with the cached copy periodically updated from the backend - One variation of this is to pre-generate the static pages as discussed above 9. Mix use all of the above if you want the best of all; consider use different programming lauguages, different application servers and different databases for different parts of a page or a site 10. ? |
|
返回顶楼 | |
发表时间:2008-12-07
csevan 写道 值得期待大牛们的方案。。。
数据库是不是也要集群啊?好像类似什么数据仓库之类的东东? 数据仓库不是数据库,根本就是完全不同的两个概念,汗一下。 |
|
返回顶楼 | |
发表时间:2008-12-23
数据仓库也来。
|
|
返回顶楼 | |