News Stay informed about the latest enterprise technology news and product updates.

Force10 creates distributed core for new data center network fabric

A new data center network fabric from Force10 Networks offers a 2 RU switch that customers can deploy in large numbers as a "distributed core."

Force10 Networks unveiled its own flavor of a data center network fabric for highly virtualized data centers and cloud computing that links together new 2 rack unit (RU) switches to form a "distributed core" and flat network architecture. The company also launched a traditional core switch and a new top-of-rack switch that holds space for embedded servers.

Force 10's flat data center switch fabric will enable up to 160 Tbps of aggregate switching capacity and as many as 24,000 server-facing 10 GbE ports.

The new Z9000 switch alone offers 2.5 Tbps of throughput with high density port capacity using either 32 40 Gigabit Ethernet (GbE) ports or 128 10 GbE ports. Once enterprises assemble a network core with multiple interconnected Z9000s they can connect that core either directly to servers or to an access layer of switches. With a distributed core of 2 RU switches, enterprises can build a flattened data center switch fabric with either one or two tiers of switching. Force10 will demonstrate the Z9000 at Interop Las Vegas in a couple weeks.

"Because of the raw performance needs of data centers today, as well as certain applications that are now being distributed in large compute clusters, compute elements need to have nonblocking performance and low-latency connectivity [with] any other compute elements. We're talking about tens of thousands or hundreds of thousands of servers that all need to be connected in a fabric, where they all have equal, high-performance connectivity," said Jeff Baher, senior director of product marketing for Force10. "That challenges traditional [three-tiered] architectures and the conventional, chassis-based systems that would be used in those configurations."

Data center network fabric: Density and manageability

Nearly every data center networking vendor is articulating its own definition and visionfor large network fabrics. In recent months, Brocade, Juniper Networks and Alcatel-Lucent have announced new products and architectures. Cisco Systems is rumored to be preparing its own fabric approach, code named Jawbreaker.

With each competing vision, vendors are trying to address two critical challenges that high-end data center managers are facing, according to Mike Spanbauer, principal analyst with Current Analysis. "One is to enable easier manageability and operational expense reductions.  The other is to have a flatter, simpler, denser environment."

With flatter, denser networks, Force10 and other vendors are trying to provide unblocked, any-to-any server connectivity with minimal latency. With a distributed core of Z9000s, an enterprise can build a data center switch fabric with just a single layer of high-speed switching with minimal hops between servers.

"It's similar in a lot of ways to what Alcatel-Lucent is doing with pods," said Zeus Kerravala, vice president and distinguished research fellow with Yankee Group. "It's a way to start building a fabric without having to invest a lot of money in a huge data center core. It provides more of a migration path to a fabric than some of the other solutions do."

Companies will be able to manage multiple Z9000s within the data center network fabric as a single device using a new feature in Force10's FTOS operating system, Unified Fabric Manager. The Z9000s will link together via Layer 2 or Layer 3, he added. The Layer 2 connections will be enabled by multilink aggregation technology and the pending IETF standard, Transparent Interconnection of Lots of Links (TRILL).

The Z9000 will be available in July at a list price of $175,000.

Server capacity in top-of-rack switches and another big chassis

Force10 also announced a new top-of-rack switch with capacity for four embedded servers within its enclosure, enabling enterprises to deploy network-integrated applications within the network switch. The S7000 is a 2 RU switch with 48 10 GbE ports and four 40 GbE uplink ports. It contains four bays where customers can slot in storage and compute resources.

"More and more companies are putting virtual servers in motion with vMotion," Kerravala said. "When you put a server in motion, all the things that surround that server—like load balancers, RAM optimizers and security— need to move with it. Now, if you want to keep [those resources] close to the application, the best way to do that might be in the network. This would give companies a platform to deploy all these soft appliances on the network."

Force10 is going to market with two partners who have developed software to run on these embedded servers: cloud orchestration vendor Joyent and security vendor Huawei-Symantec.

Force10 also introduced a more traditional data center core switch, the Z9512, a 9.6 Tbps, 19 RU switch with capacity for 480 10 GbE ports, 96 40 GbE ports or 48 100 GbE ports. The chassis will ship in 2011.

The Z9512 is aimed at customers who need between 5 and 20 terabits of aggregate switching capacity in their data center. Force10's Baher said that a traditional core switch chassis can serve that part of the market. Anything larger than that requires more ports than a chassis approach can handle. That's where the Z9000 distributed core will serve customers, he said.

The Z9512 and the S7000 will be available in the second half of 2011. Pricing information is not available.

Force10 launches an Open Cloud Networking Initiative

Along with the new devices, Force10 is aiming to create an open approach to networking, where networking pros have an ability to customize and modify networking hardware the way server and storage admins can modify their resources. So Force10 is creating a portal where customers can share scripts and software.

This "Open Cloud Networking Initiative" includes an app store, DevX Exhange, where people can share scripts and other information on the automation features built into Force10 switches.

"On the one side you have the compute culture that expects that you can open up the hood of servers and tinker with things, and on the other is the networking world which really doesn't have a hood that can be opened," Baher said. "This community portal is a way to get more people to develop around not necessarily Force10 but around networking, to share networking best practices."

Combined with the server resources embedded on the S7000, Force10 is trying to build an ecosystem of customers and third-party companies that can tinker with Force10 products and customize them to suit their specific requirements, Kerravala said.

"I think the Open Cloud Initiative, of all the solutions out there, is maybe one of the biggest risk-rewards. If they are right and can execute on this, which is tough for a small company, they could actually fundamentally change how networking is done," he said.

Let us know what you think about the story; email Shamus McGillicuddy, News Editor.

Dig Deeper on Data Center Network Infrastructure