BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Dell Networking announced a new fixed-configuration spine switch for building very large leaf-and-spine networks. Also, a new Dell OpenStack Neutron networking plug-in will orchestrate and service-chain cloud networks with help from the OpenFlow SDN protocol.
The Z9500 Fabric Switch, which starts shipping in April, packs 132 40 Gigabit Ethernet (GbE) ports into three rack units (RU). By using breakout cables, engineers can convert the Z9500 into a 528 10 GbE switch. Dell is positioning the switch as the foundation for scaling out huge spines capable of supporting a network of 100,000 servers.
"It's twice the density on a per-RU basis, compared to even the latest Nexus 9000," said Arpit Joshipura, vice president of product marketing and management for Dell Networking.
The Z9500 is based on Broadcom's Trident II network chip set, which nearly all the vendors are using in a race to come out with denser 10 GbE and 40 GbE switches, said Andre Kindness, senior analyst with Cambridge, Mass.-based Forrester Research. The switch is a bigger, faster version of the Z9000, a 32 40 GbE, fixed-configuration switch announced by Force10 Networks, shortly before it was acquired by Dell.
Dell has priced the Z9500 in a way that mirrors chassis-based systems, where engineers add line cards when more ports are needed. Engineers can order and license the Z9500 based on three SKUs, with 36, 84 or 132 active ports. As an enterprise's network grows, engineers can activate additional ports through a licensing mechanism.
"You have this low-cost race going on in on the [network] hardware side of the market," Kindness said. "This is another way to approach it." Brocade has offered pay-as-you-go pricing on its switches for more than a year. With Dell embracing the concept on the Z9500, other networking vendors will follow, Kindness said.
Active Fabric Controller: A Dell OpenStack Neutron plug-in
Dell's Active Fabric Controller (AFC), meantime, is an OpenStack plug-in that uses the OpenFlow SDN protocol with proprietary Dell extensions to automate configurations and service chains for new workloads provisioned in OpenStack. The AFC will ship this quarter.
More on Dell Networking
Dell partners with Cumulus for bare-metal switches
A new generation of Dell campus switches
"If OpenStack workload Y wants you to dynamically stitch the fabric in one way and apply security policies and require Layer 4-7 services to it, all of that can be done in a seamless, automated manner using AFC," Dell's Joshipura said. "As you put VMs and workloads into OpenStack, AFC -- using extensions to OpenFlow -- [discovers] the switches and [builds] a topology map. Now you have this topology map without running BGP [Border Gateway Protocol] or a spanning tree. You can push that map down to the switches using extensions of OpenFlow. The switches now have tables that move the packets appropriately."
Although AFC uses OpenFlow, Joshipura declined to describe it as an SDN controller.
That's an appropriate distinction, said Christian Renaud, senior analyst with New York-based 451 Research. "If you think about how Big Switch or NEC uses OpenFlow, where the controller is in the packet path, [OpenFlow] is really involved and much more interventionist," he said. "[AFC] is different. It is only using [OpenFlow] as a signaling protocol. Traditional switches are receiving policy messages from [AFC]."
Dell also announced that it will partner with Red Hat to develop AFC as the foundation of a network functions virtualization (NFV) platform. AFC's ability to service-chain networks, security and Layer 4-7 services will help service providers orchestrate and chain together virtual network functions in their data centers, Joshipura said. The NFV technology will ship some time in 2014, he said.