KODO JDO分布式缓存性能如何?
有谁有KODO JDO的分布式缓存机制的经验吗?我想知道:
1)分布式缓存更新之间的延迟是多少(因此,如果两个用户访问两个单独的缓存,即在两个不同的 JVM 上,并使用相同的数据,其中一个用户进行更新,那么另一个用户何时会进行更新,使用其他缓存,看到更新了吗?)
2) JVM 之间将传输多少数据?如果对一个缓存进行更新,它是否只是通过告诉其他缓存要刷新的对象的主键来告诉其他缓存删除对象? (关注的是管理分布式缓存的网络流量/开销)
3)当您全天有外部源更新数据库时(即不是通过您的应用程序进入),从外部调用缓存刷新有多容易?
我们的应用程序在 12 个 JVM 的 Weblogic 集群中运行,我们正在考虑启用分布式缓存,以帮助提高从数据库中提取的大型对象图的性能(目前尚未缓存),但我们想了解一些真实情况#1、2 和 3 的经验。谢谢。
Does anyone have experience with KODO JDO's distributed cache mechanism? I would like to know:
1) what is the latency like between distributed cache updates (so if two users are hitting two separate caches i.e. on two different JVMs and are using the same data and one makes an update, when will the other user, using the other cache, see the update?)
2) how much data will be transferred between JVMs? if an update is made to one cache, does it simply tell the other caches to drop the objects by telling it the primary keys of the objects to flush? (concern is the network traffic/overhead of managing the distributed cache)
3) when you have external feeds updating your database throughout the day (i.e. not coming in through your application), how easy is it to externally invoke a cache flush?
Our application runs in a Weblogic cluster of 12 JVMS and we are considering enabling the distributed cache to help with performance coming from large object graphs being pulled from our database -- which are currently not cached-- but would like to know some real-world experience with #1,2,and 3. Thanks.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
这是部分答案,但我相信仍然有帮助(来自 http:// /docs.oracle.com/cd/E13189_01/kodo/docs303/ref_guide_cache.html):
没有说明这是否意味着此提交被包含为原始事务的一部分(人们希望如此)或和/或此操作的延迟时间或开销以及它的扩展程度(例如,如果您执行此操作,它会如何执行)正在协调 15 个 JVM,并且有多个用户更新相同的数据)
This is a partial answer, but I believe still helpful (From http://docs.oracle.com/cd/E13189_01/kodo/docs303/ref_guide_cache.html):
It is not stated whether this means that this commit is included as part of the original transaction (one would hope) or and/or what the lag time or overhead is with this operation and how well it scales (e.g. how does it perform if you're coordinating 15 JVMs and you have multiple users updating the same data)