开发者

Communication between 2 nodes in a cluster

开发者 https://www.devze.com 2023-02-10 16:52 出处:网络
I am trying to figure out how this will work out: client-server communication via NIO/BIO Server-server communication (replication, membership etc) via JGroups (replication of data can be a pain?)

I am trying to figure out how this will work out:

  1. client-server communication via NIO/BIO
  2. Server-server communication (replication, membership etc) via JGroups (replication of data can be a pain?)

What are your suggestion, pros/cons of using such an architecture, why go for jgroups, any alternatives?

Should I go with native NIO or use some libraries like Mina or Netty?

Update

  1. No REST, it's not a web server. I am buildin开发者_运维问答g a dynamo like distributed database.
  2. activeMQ is like JMS I think, where I will have single point of failure. JGroups solves that issue, right?


I wonder why you you want to implement binary protocol for client-server communication? My advice is to go with REST or WebServices. It has huge advantages compared to proprietary binary protocol.

About server-server communication... There are huge number of choices and I wonder why you have chosen JGroups. Can you elaborate on this? I will advice you to go with some messaging solution like JMS. There are a lot of available implementation like ActiveMQ. It will nicely decouple servers from each other.

If you need to maintain shared configuration information, naming, have distributed synchronization between servers you can look at Apache ZooKeeper project.

If your servers are deployed at Amazon EC2 instances, than I think the best solution is to use SQS and/or SNS services for server-server communication.

You can use Terracotta for replication of your server's state (still I will recommend you to make your servers as stateless as possible).

But it's really hard to advice anything more concrete because you have not described neither your system nor requirements to it. I think it would be very helpful!

0

精彩评论

暂无评论...
验证码 换一张
取 消

关注公众号