Appliance wars -- In-band or out-of-band
Today's SAN virtualization startups provide solutions for managing and virtualizing storage from the network but they don't seem to be implementing caches on these switches. How are they coping with latency issues and how much of an effect does this have on performance of these systems? Having to route write operations all the way to the end storage systems will add significant overhead on these operations.
Also, IP SANs are coming so what advantages/disadvantages do you see by moving storage access to IP networks? I have been in the storage industry for 11 years and to me it seems that IP SAN will resolve a lot of the issues that exists in FC-SANs such as cost, security, WAN access and management. What are your thoughts on this?
Your assumptions are valid in my humble opinion. There is almost a holy war occurring between proponents of appliance based in-band virtualization and fabric, network or metadata appliance based out-of-band replication technologies.
I think the in-band solutions approach is a great way to get storage pooling across multi-vendor subsystems, in a cost effective manner. Although I don't think it may be as scalable as the out-of-band approach, it's here today and can be a cost effective solution to provide data replication between different storage devices, thus reducing the overall costs of a disaster recovery solution.
I still personally like the out-of-band method for it's scalability.
IP SANs are not coming, they are here now. I currently have customers using the technology to great advantage. iSCSI makes for a great solution for inexpensive disaster recovery mirroring and consolidated backup from remote offices. It also is great as a tool to re-use existing Gigabit Ethernet networking gear as a SAN backbone.
Pure iSCSI SANs are a bit farther off. Enterprise class iSCSI storage is starting but there are limited products out there. This will change by the end of 2003. There are plenty of TOE adapters coming out on the market so the host end is almost done. We now need to see iSNS services and more comprehensive security take off in the fabric layers of both SANs and IP networks for it to really take off.
When we start seeing native iSCSI support at the operating system kernel layer, which is coming, then things REALLY start to change.
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our
This was first published in January 2003