On Fri, May 14, 2010 at 4:27 AM, Dot Yet <dot....@gmail.com> wrote: > Hi Everyone, > > I am trying to setup a small size iscsi based storage. Total 16 > spindles in a system, connected to two Solaris 10 hosts through intel > gig nics. The disks are configured in a mirror, and locally they > perform OK. When I try to export them through COMSTAR to those Solaris > 10 hosts, unfortunately, the performance tanks. I can only get around > 70 to 80 MB/s. all the nics running iscsi traffic are connected to the > same unmanaged switch. > > I have read about link aggregation. Do you think adding a quad port > intel NIC and a managed switch can help in getting higher speed over > iscsi. I also read that having a single connection to the target will > always use only one of the linked nics, not more than one, so one will > stay limited to 1 gb throughput only. Right now, I have 4 connections > per target. do you think link aggregation will spread these 4 > connections over the 4 nics and give me higher throughput. I am not > looking for exact 4 gb wire speeds, but still something better than 1 > gb. > > Also, do i really need a managed switch to connect these or can this > be done through a direct cross cable connection? > > thank you in advance. > > d.y > _______________________________________________ > storage-discuss mailing list > storage-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/storage-discuss >
You will probably want to look into iSCSI MPIO on your initiator side. I am not as familiar with the Solaris iSCSI initiator, but most OS's have MPIO support built in for link load balancing based on various algorithims, like disk queue depth, round robin, failover only, etc -- Brent Jones br...@servuhome.net _______________________________________________ storage-discuss mailing list storage-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/storage-discuss
