What network utilisation can you get from 1gigabit?

I am building a cluster with 1 gigabit cards (actually pairs of HP cards working as 2gigabit ‘teams’) but when replicating stuff I can’t get much above 2% utilisation which is a bit dissapointing. I can’t get much better copying files across the windows network either. Both servers are running windows and on the same gigabit switch. any ideas on what is holding it back? or is this all I should expect?

Subject: what network utilisation can you get from 1gigabit?

You’ll probably only increase bandwidth utilization if (a) you have multiple threads/processes transferring data simultaneously, and (b) your i/o subsystem can keep up with the read and write requests from the various threads.

Subject: try ftp?

Alan, in my experience ftp has been the best method for transfering large files in the least time, it would be interesting to see if that works for you:-jmw

Subject: what network utilisation can you get from 1gigabit?

I have built many Domino clusters (using private switched network, crossover cables… for intracluster LAN). The bottlenecks have not been the NICs (network bandwidth…) but the ratio of “volume of data : cluster configuration”. What is cluster replication queue? If there are backups, you might increase the # of cluster replicators. I would be cautious re: adding too mant clrepl tasks. The cluster configuration recommendation or consideration is made based on the assumption that the hardware is capable of handling the load. Needless to say, your cluster replication process is tied to server performance. It can also be the little (not so little) things like having a robust L2 cache on the server. Domino uses L2 cache on most operations.

Subject: RE: what network utilisation can you get from 1gigabit?

I don’t think it is actually to do with the clustering. At the moment I am doing accelerated create replicas between the two, I tried a script to walk down a notesdbdirectory and do notesdatabase.createreplica to build the replicas on the empty server. I think this is all traditional replication between the boxes. I wouldn’t expect the cluster traffic to be that big, I am just trying to fill up about 150Gb of stuff over the network, in theory that should be a minimum of 150*8 seconds=20 minutes, at 100% utilisation which is of course impossible, however at 1% it is 33 hours. I was figuring it would do about 20% and be done in a couple of hours.

Subject: RE: what network utilisation can you get from 1gigabit?

thanks for the tips,

sh stat replica.cluster.workqueuedepth

Replica.Cluster.WorkQueueDepth = 0

1 statistics found