Nexsan SASBoy

For our host systems we used Boston Supermicro dual Xeon 5100 and Dell PowerEdge 1950 dual Xeon 5300 servers both running Windows Server 2003 R2. Each was fitted with LSI 4Gbps FC HBAs and ran the latest Microsoft iSCSI initiator. Once FC connections were established we could see each host’s WWPN in the SASBoy’s web interface and could then dish out read/write access for them. The same applied to iSCSI as once the initiator had been logged in to the relevant target port its IQN was also displayed.

Performance over FC is impressive with the Iometer utility reporting a healthy raw read rate for one server of 366MB/sec – pretty much wirespeed for 4Gbps FC connections. With the second server also running Iometer on its own volume we saw a high cumulative read throughput of 731MB/sec. To get the best IP SAN performance we found you absolutely must use Gigabit jumbo frames. Without this configured we saw our two hosts return read speeds of only 54MB/sec apiece. Using a Netgear GSM7328S switch, we configured support for an MTU of 9216 bytes and activated support on our two servers. With jumbo frames rampaging across the network we saw each host now delivering between 94-97MB/sec raw read rates.

The SASBoy supports iSCSI’s MPIO feature and any host that uses Microsoft’s iSCSI initiator v2 and above will have the necessary DSM (device specific module) as standard. This lets you create redundant paths to storage volumes by allowing Windows to see the same disk twice so paths from two network controllers to the same logical drive can co-exist. For the SASBoy, we mapped one volume to both data ports on one controller and placed them in different subnets. We created a failover link using both network ports on our Dell server and with Iometer fired up we were able to disconnect one network cable and watch it continue unabated.

The AutoMAID feature worked very well during testing and we set the head parking time at two minutes and spin down at fifteen minutes. We were unable to select the second option for reducing spin speed as the Hitachi disks don’t support this feature. Nevertheless, after the first option came into force we saw consumption drop from 397W to 365W and after spin down this fell further to only 224W. When accessing the array from one our test hosts we could see no appreciable lag when the drives powered up ready for action.

To create the SASBoy, Nexsan has simply taken a SATABoy 2 system and replaced its drives with SAS models. These certainly made their presence felt in our performance tests and we found the AutoMAID feature can make big power savings whilst performance across the board is good.