11
Looking Ahead
This paper discussed setting up iSCSI on 1Gb/s Ethernet to compare performance of
running an iSCSI initiator in software versus using an HBA. The most obvious next step
would be to test both on 10Gb/s Ethernet. My expectation is that an HBA will be required
to see any performance gain.
Although network and host performance optimizations were not investigated, it is
unlikely to be worth the time spent (for most practical uses) given that the limiting
performance factor is the network speed.
It is very interesting that the average speed for both Windows tests was 42MB/s. Once
again, this raises the question of whether the HBA or the software initiator was being
used. It’s hard to tell because after installing the HBA driver, the Windows iSCSI control
panel picked up all the HBA settings and automatically logged into the router with these
new settings. It might be interesting to build a fresh box and install just the HBA (not the
software initiator) and run the test again. However, given the lack of significant
performance gain on the Linux box, any change on the Windows re-test is not expected.
Finally, other storage technologies utilizing Ethernet (such as hyperSCSI,
ATAoverEthernet and SATAoverEthernet) should be investigated and compared for
performance, cost, setup requirements and scalability.
