Cisco expanded and refreshed its Nexus data center switching catalog, announcing a smaller Nexus 7700 switch, a...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
higher-density Nexus 5600 switch and a copper Nexus 3100 switch. The company also introduced port modules for the Nexus 7700 and the Nexus 6000 and added KVM hypervisor support to its Nexus 1000V distributed virtual switch.
With these new products, the Nexus product family is becoming extremely large, with three series of switches recommended for the core (Nexus 7000, 7700 and 9000), four series recommended for the aggregation and spine layers (Nexus 6000, 7000, 7700 and 9000) and at least five series recommended for access and leaf layers (Nexus 2000 fabric extenders, Nexus 3100, Nexus 3500, Nexus 5600 and Nexus 9000).
"It's a whole bunch of products, and that can sometimes be confusing for field sales organizations and channel partners," said Christian Renaud, senior analyst with New York-based 451 Research. "If you are a value-added reseller looking at this, you might be saying this looks like alphabet soup."
What looks like alphabet soup to the sales channel will present an even more confusing picture to engineers trying to decide what path to follow.
At the same time, the sheer volume of Cisco's customer base demands this breadth of data center switching options, Renaud said. "What other manufacturer has the scale to be able to say that we have 5,000 customers who want a 40 Gb pizza-box switch, so let's just build one. The upside is they have point products for point applications."
In order to help customers wade through its various product families, Cisco is organizing its Nexus line into seven use cases and recommending an architecture tailored to each one. Those use cases include private and hybrid cloud networking, big data, high-frequency trading, LAN-SAN convergence, Web 2.0 and massively scalable data centers, business continuity and data center interconnections, and general enterprise applications. An engineer can pick a use case and Cisco will offer an architecture that mixes and matches all of these switches.
"This is emblematic of where we are in the transition of the networking industry," said Brad Casemore, research director for Framingham, Mass.-based IDC. There isn't a single data center network architecture that fits every use case because the industry isn't moving in the same direction or at the same pace. "So you're seeing fragmentation of solutions across the market. You see it more with Cisco because they cover the whole market. It's a blessing and a curse to be the 800-pound gorilla in networking. You have customers that run the gamut from small and medium business to large enterprises to cloud service providers to people trying to do high-frequency trading. You have to provide architectural frameworks for all of them."
Many of the switches announced this week are also examples of Cisco shutting off potential avenues of attack from competitors by offering switches right-sized for every scenario, Casemore said.
A junior Nexus 7700
The Nexus 7706 is a prime example of that right-sizing. The new chassis switch is a baby brother to Cisco's biggest family of data center switches. At just nine rack units (RU), the Nexus 7706 is much smaller than the 26 RU Nexus 7718 and 14 RU Nexus 7710. The Nexus 7706 offers 21 Tbps of switching capacity and a maximum of 192 by 10, 96 by 40 or 48 by 100 Gigabit Ethernet (GbE) ports.
In addition, Cisco released new member of its F3 series of line cards for the Nexus 7700 chassis, a 48-port 1 or 10 GbE module that supports onePK, Cisco's application programming interface that enables programmability in its switches.
"[Cisco] is trying to get that midmarket core [niche] addressed," Casemore said.
Renaud said 451 Research's surveys have revealed that a large number of enterprises are preparing to upgrade their cores in 2014. Cisco is targeting enterprises that are looking to leapfrog from the Catalyst 6500 to the Nexus 7700, skipping right past the Nexus 7000 series.
Nexus 5000 family refresh: Bigger and better with VXLAN
The Nexus 5000 family received a refresh with the Nexus 5672UP and Nexus 56128P. Like their predecessors, the Nexus 5600s can serve as server access switches, or -- when deployed with Nexus 2000 fabric extenders -- as aggregation switches. The Nexus 56128P pushes the 5000 series' 10 GbE port density to 128 with additional 12 by 40 GbE ports. Both switches have Unified Ports that can operate in Ethernet, Fibre Channel or Fiber Channel over Ethernet mode. The switches also support OpenFlow 1.0, onePK and in-hardware VXLAN routing and bridging.
The Nexus 6004, Cisco's semi-fixed-configuration spine switch with high-density 10 and 40 GbE, will also get Unified Port support thanks to a new 20-port line card designed for its expansion slots.
Nexus 3100 with VXLAN support in copper 10 GbE
Cisco expanded its line of ultra-low latency switches with the Nexus 3172TQ, a 10 GbE copper switch, the first copper switch to support VXLAN, according to Shashi Kiran, Cisco's senior director, market management for data center, cloud and open networking. "We see this having resonance in massively scalable data centers and as a general purpose, top-of-rack switch," he said.
The Nexus 3172TQ is a "Swiss Army knife" platform, Renaud said. Engineers can expect Cisco sales to push this switch on them as a versatile access layer device. Because it supports copper rather than fiber, engineers will be able to upgrade to higher bandwidths without refreshing cables.
KVM support for Nexus 1000V
Finally, the Nexus 1000V distributed virtual switch added support for the KVM hypervisor. In addition to VMware and Microsoft Hyper-V support, the 1000V can now cover a wide range of multi-hypervisor data centers, Kiran said.
"IT's not just about having it integrated with hypervisors," he added. "There is also a lot of other functionality -- security services, WAN optimization, linkages to hybrid cloud frameworks, orchestration platforms. All of them become consistent, regardless of which hypervisor stack is being run."