Long distance application failover between data centers.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-07-2006 07:07 AM
Hello:
I am working a project where there is a requirement to have a primary server and a secondary for application failover. The requirements dictate that the servers must be geographically separate at separate data centers but yet on the same IP subnet so that heartbeat messages are not natively routed. What protocol can I use to bridge the subnet over an IP routed network? GRE? L2TP?
- Labels:
-
Server Networking
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-07-2006 07:54 AM
GRE will not work as according to the last information I have you cannot bridge over a GRE tunnel. We had one such application with similar requirements and we ended up using L2TPv3 that allows transparent LAN extension. We have not had any issues with it and if you are only going to use it for heartbeats you should not run into any performance issues. L2TPv3 does tend to drive the CPU utilization high in case you have a lot of data traversing the tunnel.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-10-2006 09:31 AM
That type of failover is not really scaleable between datacenters.
If it is linux HA you can eliminate the same subnet problem by using multicast and ping between the two nodes and bridging or routing that mutlicast traffic between the two networks.
