In order to get the most out of next-generation data center networking technologies such as Fibre Channel over...
Ethernet (FCoE), enterprise IT departments must not only conquer interoperability issues, but also break down organizational silos within the data center.
That was the view of some of the world's leading network equipment vendors expressed during an Interop keynote panel Wednesday titled "Reinventing the Data Center."
A case in point for the panelists was the gradual movement toward converged connectivity in the data center, described as data center Ethernet or converged Ethernet. One key element of that convergence is FCoE, which offers data centers the ability to consolidate Fibre Channel storage traffic onto Ethernet infrastructure. This technology can reduce cabling and the number of network adapters connected to each server, which cuts down on management overhead and energy consumption.
"You have servers with four or five Ethernet adapters and a couple of HBAs that go off into disparate networks, and there is some cost savings in taking those adapters and combining them together and aggregating that value into the network," said Dave Stevens, chief technology officer at Brocade Communications Systems. "But I don't think we can underestimate how tough that is to do in an organization where there are four or five Ethernet adapters on a server today."
Stevens said there is no technical barrier to consolidating those adapters, but nobody does it. One reason for a lack of movement on consolidation is the siloed nature of IT organizations within the data center. He said no one can agree who would own service-level agreements or the provisioning, for instance, on those consolidated adapters. On a converged network, if the storage network goes down, IT organizations need to know whether to call the networking team or the storage team.
"Storage and networking have got to blend those organizations together and blend the technologies together," Stevens said.
The siloed nature of data center IT organizations also contributes to slow application deployment cycles, according to Paul McNab, Cisco's vice president for enterprise and midmarket solutions marketing.
"That is the problem you have today with silos, where it takes three to six months to deploy new applications and servers, because I've got someone managing MAC addresses. I've got my network guy, my storage guy, my compute guy," McNab said.
The sponsor for a new application needs to check in with all these different teams before launching a new service, he said. And negotiating who owns the processes for provisioning and managing the IT infrastructure to support those applications is a maze.
"One of the big things we have been working on is how you can create a service profile, where you can have each of those different silos keep their domain expertise and policies, but maybe you integrate them into one embedded management capability," McNab said. "So you can just cut and paste those service profiles that cover the servers and the network and the virtualization policy into one integrated policy. And when you want to dynamically turn on another compute node, you can just turn it on automatically."
McNab hypothesized that 70% of an IT budget is spent on managing the siloed data center. He suggested that an integrated system that bridges silos would cut not only time to deployment, but costs as well.
David Yen, executive vice president and general manager of the data center business group at Juniper Networks, said networks are the foundation that will enable the collapse of silos in the data center.
"The stars in the data center are servers and storage," he said. "But the data center network is the foundation underlying the whole data center, so it does have to cooperate with server orchestration platform providers."
Yen said the network needs to integrate fully with the system management technologies that enable the dynamic provisioning of server and storage capacity, but networks must also be transparent so the rest of the IT organization doesn't need to deal with the complexity of the network.
"Precisely which physical port a server plugs into is irrelevant as far as the end user is concerned," he said.
Although the representatives of the various vendors concurred on the importance of breaking down silos in the data center, Stevens and Yen also managed to fire a couple shots at their colleague from Cisco. For instance, they indirectly criticized Cisco's entry into the blade server market and the company's penchant for developing proprietary technology that isn't always sympathetic to the open, standards-based approach many other vendors try to champion.
"We think the vertical integration into the server space will provide short-term facilitation for the customers, but in the long term it actually deprives customers the long-term benefits of a best-of-breed evolution that is happening in each component of the [data center] industry -- and proprietary lock-in will not be beneficial for customers," Yen said.
Stevens added, "We agree that [storage, servers and networks] have become tighter over the years, but the data center is by definition a silo problem. The data center that evolves will have to incorporate all these technologies in an open, standards-based way that gives customers the flexibility to pull components out and put other components in."
McNab defended Cisco's record by arguing that standards bodies aren't "innovation bodies." He said vendors should strive to differentiate their technology ahead of any standards that are ratified. "We need to improve interoperability to the point where we can all have some differentiation. We must try to ensure that customers get the benefits of plugging these [standards] in and that they are able to integrate far better."
Let us know what you think about the story; email: Shamus McGillicuddy, News Editor