Routing iSCSI with 10GIG

We recently redesigned our DC network templates. Part of this redesign created a “Storage Pod” where our network storage resources would be located and virtualized. This consolidated storage devices not just physically, but logically into separate VLANs. The previous DC design had the storage resources in the same VLAN as the application and database servers. All iSCSI access was at Layer-2, ignoring the “i” in iSCSI. This was a legacy decision driven by our storage teams who were convinced that any Layer-3 routing of iSCSI would severely impact performance. Nothing could be further from the truth with Layer-3 switching and ASICs, but Layer-2 connectivity for storage access was the SOP. We broke out of this concept with the new DC design, showing the storage and hosting teams that iSCSI access at Layer-3 was just as fast, if not faster, than at Layer-2. The driving technology behind this decision was 10GIG Ethernet. 10GIG gave us the bandwidth and connectivity needed to multiplex our storage data flows over a Layer-3 network into the Storage Pod. Instead of each iSCSI storage head having interfaces to every application VLAN, now the heads have two 10GIG interfaces and route traffic via the network to the servers. This reduces cabling, port utilizations, and power. Plus, we get the benefits of Layer-3 routing (simpler security, load balancing, traffic identification, etc). Before 10GIG this design was not an option, particularly from a performance standpoint. 1 Gbps interfaces, even several of them, did not provide the bandwidth needed for multiplexing intensive application and database traffic. A couple 10GIG interfaces does. Testing and early deployments have been positive. 10GIG made it possible to virtualize our network storage and optimize our data center design. 10GIG is changing the data center in more than one way – it’s not just more bandwidth.