[Gluster-users] rebalance speedup

[Shishir Gowda]

Hi ling,

The feature for speeding up rebalance by allowing multiple servers to support parallel rebalance is in release 3.3.0.

A single rebalance processes is started on every node/server where the volume's bricks exist.
These rebalance processes share load between themselves to migrate data, hence reducing the bottleneck of having only 1 rebalance thread do the work.
Additionally, all rebalance processes push data to the destination node, hence better utilizing the network too.

Hope this answers your query.

With regards,
Shishir

----- Original Message -----
From: "Ling Ho" <ling at slac.stanford.edu>
To: Gluster-users at gluster.org
Sent: Wednesday, June 13, 2012 7:02:30 AM
Subject: [Gluster-users] rebalance speedup

I found a discussion earlier this year regarding a planned feature for 
speeding up rebalancing in 3.3.0.

Is these a way to speed up rebalancing now in 3.3.0? Is it possible to 
let rebalancing between multiple servers making use of more of available 
network bandwidth?

I have 5 servers with 10Gb network, and 40Gb infiniband. The copying is 
going on the 10Gb interface only when I started rebalance, and it's 
consuming less than 1Gb/s for each streams.

Thanks,

...
ling
_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://gluster.org/cgi-bin/mailman/listinfo/gluster-users