The following post was authored by Gary Gumanow (Product Marketing Senior Consultant, EqualLogic).
At Interop Fall 2010 Dell is demonstrating the latest in Ethernet technology, Data Center Bridging (DCB), which are a recently ratified IEEE (eye-triple-e) group of standards to ensure quality of service on data center Ethernet networks. This demonstration will show the benefits of what DCB can do in a converged network environment. With DCB, Ethernet finally delivers on the promise of network convergence or unifying all types of network traffic. It can do so by shaping bandwidth for particular flows of traffic with Enhanced Transmission Selection, for such applications as inter-processor communication, or storage.
Another enhancement to Ethernet that we are showing is Priority Flow Control, or PFC. This part of the demo will showcase DCB's ability to setup different classes of network traffic. You know, all types of traffic were not created equally. Ethernet has had prioritization for almost ten years, but it's not had the ability to prioritize a specific type of traffic. Halt the LAN traffic so that storage traffic gets through. That's the idea of PFC, and we'll be showcasing this.
To make this all work smoothly, DCB includes an "auto-negotiation" type protocol, which acts as a handshake mechanism to automatically determine its partner's capabilities. "Can you do PFC, ETS?" We'll be showing this capability as well.
As you may know. iSCSI has an inherent mechanism to protect against losing, or dropping, packets. It's called TCP/IP. It's been around for a long time. Well, DCB was created so that Ethernet would not drop a packet. So if iSCSI uses TCP/IP and doesn't drop a packet, why does it need DCB? Well, DCB was really created for Fibre Channel over Ethernet or FCoE, because Fibre Channel cannot drop a packet... it has no way to recover in a neat and timely fashion. So, do I need DCB for iSCSI? The short answer is... it depends. Here's a link to a blog I wrote that speaks to this. Are you dropping a lot of packets? If so, DCB can eliminate retransmits, thus making TCP/IP more efficient. It might even reduce your CPU utilization because of this. And you could use those cycles to virtualize more of your data center.
And lastly, DCB is an open IEEE standard. Dell's solution and approach is of course open. We are working with partners to ensure that our customers have the widest selection of options. For this demonstration, we'll be showing Brocade switching, Intel network adapters, Dell PowerEdge servers, and EqualLogic PS-series storage targets. We've been working hard over the last two years to make sure we have the most robust solution in the industry, and working with industry alliances, such as, the Ethernet Alliance, and the Storage Networking Industry Association, and UNH-IOL Labs. Also, keep in mind that DCB is a 10GbE technology. It only works today with 10GbE. Here is a SNIA webcast we did on how to "Optimize Your Storage Enviroment With iSCSI SANs".
I hope to see you at Interop NYC. Please stop by and say hello.