Does 10GigE + iSCSI = FC is in trouble???
Posted in General, Storage Interconnects & RAID, Advisor - Joe Disher by Joe DisherAt this weeks Storage Networking World conference in Orlando, Florida an IT Manager asked me what I thought of 10 Gigabit Ethernet and it’s effects on Fibre Channel. “Especially with the growing popularity and acceptance of iSCSI,” he added.
Clearly he had been talking with one of the other vendors on the exhibition floor and had some thoughts of his own that were sparked by this vendor’s view of the situation. I gave the gentleman my view on the topic and thought it would be an interesting topic to talk about today.
BTW, back in early September, Marc Staimer wrote a column on SearchStorage on this topic that I feel clearly spells out the realities of this debate.
Here’s the summary of what I told that IT manager - you’ll find that Marc and I tend to agree on this subject.
For me, I think it boils down to a few key pieces:
1) Final ratification of the 10GigE standard for Cat6 and Cat7 copper cabling
2) Availability of 10GigE copper ports on switches and iSCSI target devices that support 10GigE
3) Cost per port
Once the Cat6 and Cat7 standards get nailed down for 10 gigabit copper the road for development of iSCSI target devices supporting 10GigE will finally be paved.
The Ethernet switch vendors are more than ready to respond to having ports available to be plugged into their switches. They clearly have a vested interest in making sure the standard is completed quickly and that they are ready with products once it’s done. There’s one big name Ethernet switch company that would love to steal a big piece of the “SAN” pie from the FC guys. The other natural follow on to the above mentioned standards being completed will be the fast introduction of iSCSI target devices supporting 10GigE copper connections. Again, there is lots of interest from the iSCSI supporters to have an advantage over FC.
Finally and maybe the most important piece of them all, the day that the cost per port for 10GigE copper gets down to anywhere near or (gasp!) maybe even less than 4Gb FC, the time for 10 Gigabit Ethernet will finally have come. There’s no denying that customers feel more comfortable with the ubiquitous Ethernet technologies than they do with the mysterious and complex FC SAN technologies. So once prices come down and FC no longer has a technological advantage… how could anyone deny the unbelievable potential for iSCSI and 10GigE. So does 10GigE + iSCSI = FC is in trouble??? Not today, but it will soon! Look for the transition to start happening in 2006.
Let me know how you view the chances for 10GigE.
Joe
March 7th, 2006 at 3:13 am
this point of view is mainly focused on $$, obviously very important.
but what about performace, is kwown that tcp/ip stack overload on servers is greater than fcp, this is negligible when you’re transferring small amounts of data.
in case of high transfers (i.e. 500 MB/s), your server CPU will suffer and perhaps you’ll need to increase the server CPU power, adding cost to the solution. And this for every server.
April 15th, 2006 at 4:54 pm
ISCSI offers great performance - but it requires much more tuning. Adjusting your TCP window size, tuning NIC parameters, configuring your network to support frames larger than the standard 1500 byte frame - these are all important when looking to configure ISCSI for performance.
These techniques combined with the use of offload engines are sure to grant the implementor a successful deployment