Cisco's switch upgrades tout speed, ease, innovation

Cisco Systems has introduced a new chassis and supervisor module in its midrange Catalyst 4500 series of switches, and a new supervisor for its flagship Catalyst 6500 series. Both sets of announcements -- one evolutionary and one revolutionary -- give Cisco admins and network architects plenty to get excited about..

The new Catalyst 4500 E-series delivers a substantial speed boost and offers an easy transition from 1Gbps to 10Gig networking. The new Supervisor 720-10G module for the Catalyst 6500 series combines the obligatory increase in throughput with a whole new approach to redundant networking called the Virtual Switching System. These enhancements in the 6500 series switches are sure to change many network designs in the coming months and years.

Cisco also announced the Cisco SMART Call Home Service, a service that allows Catalyst 6500 series switches to push diagnostic information and service data back to Cisco. Call Home can be used to generate automated service calls and produce Web reports for admins to gauge the performance of their infrastructure.

The new Supervisor 720-10G brings the Catalyst 6500 up to a total throughput of 1.44 terabits per second, and runs 40Gbps full-duplex to each linecard. It's also the brains behind Cisco's intriguing new switching concepts.

Virtual Switching System (VSS) looks to simplify redundant networking designs significantly, and do away with the love/hate relationship and power struggles network admins have with switching protocols such as Spanning Tree Protocol (STP) and HSRP/VRRP (Hot Standby Router Protocol/Virtual Routing Redundancy Protocol).

In current networks, true redundancy at the core and middle tier is achieved by careful planning, careful configuration, and a healthy dose of STP and HSRP. STP provides the smarts necessary to determine what path a packet will take to its destination based on link availability and constant communication with neighboring switches. HSRP and VRRP allow multiple independent layer-three switches to cooperate on providing route availability, with the routing engine of one switch taking over upon a failure event of another. In large networks, these protocols are ubiquitous and sometimes notorious. Poor configuration of these protocols can cause networks to be unnecessarily slow, or even crash. In a properly functioning network, STP reconvergence and port inspection can take up to 50 seconds, and HSRP/VRRP problems can easily result in network downtime.

Now Cisco is eliminating these protocols with VSS and the Supervisor 720-10G. In a redundant core configuration, for instance, two Catalyst 6500s operate as a single unit, sharing a single management address and a single configuration, and cojoined by the 10G interfaces in each supervisor. This isn't really stacking in the normal sense of the word; rather, it seems to be more of a marriage between two cores, tying them so close together they share address, switching, and routing tables. VSS has lots of apparent benefits: Core reconvergence times should be greatly reduced, multiple cores can be managed from a single interface, and the headaches associated with HSRP and STP go away.

Page Break

Pushing out from the core, Cisco's middle-tier gear plays this game as well, with all switches handling best-path calculations automatically, which should reduce configuration complexity as well as latency. It will be interesting to see this in practice, since changes of this magnitude at the network core don't happen often. One caveat is that redundant supervisors aren't yet supported with VSS, so each switch can only run a single supervisor engine. I'm also curious to see how this works with access-layer switches that still need to play by STP rules.

Multi-chassis EtherChannel has eluded Cisco until now, but it's a reality with the Supervisor 720-10G. This provides very significant benefits in highly available environments. Previously, server network redundancy required a sacrifice -- either you could have redundant links to two switches, or you could bundle those links for increased bandwidth, but link them to only one switch. With VSS in the Supervisor 720-10G, now you can have both increased bandwidth and redundant links.

The new Catalyst 4500 series chassis, dubbed the 4500 E-Series, carries with it a significant throughput bump, bringing each slot up to 24Gbps from 6Gbps, as long as you're using the new line cards and Supervisor 6-E. Existing line cards will work in the chassis, but will be limited to their original 6Gbps speeds. This does not impact other slots on the chassis, however, so you can mix and match older and newer line cards without impacting the performance of the new hardware. These new cards include a 48-port 10/100/1000 module with PoE standard on all ports. The premium version of the same module carries the ability to run 30 watts per port to any 24 ports on the card, which can be necessary to drive high-powered PoE devices, like wireless access points running several radios.

Also new is the 6-port 10G line card, that conforms to Cisco's TwinGig specification. TwinGig is designed as a stepping stone to full 10G deployment. Using a TwinGig adapter in a 10G slot enables two independent 1G fiber links to be driven from the 10G port. Down the road, 10G optics can be used in the same slot to deliver a single 10G port. Essentially, this 6-port 10G blade can become a 12-port gigabit blade with a clear path to 10G without any major upgrades.

The new supervisor module for the 4500 E-series, the Supervisor 6-E, provides the horsepower to drive each new line card at 24Gbps, but delivers more than just a speed boost. For one, it incorporates both USB A and USB B connections on the front of the card, allowing quick access for flash devices and providing a method for admins to directly address the internal flash storage of the supervisor.

Anyone that's ever spent too much time trying to upload OS images to a supervisor via a serial cable and Xmodem will appreciate this addition. Also, the 6-E has two TwinGig-capable 10G ports, and an Ethernet management port.

The Supervisor 6-E has a few nice internal features as well. For instance, the requisite IOS image on the 6-E eliminates hard barriers such as memory storage limits on internal ACLs and QoS rules, a welcome addition to those who are running 4500s in complex networks with large security and QoS requirements. Also, when used in the 4507R and 4510 chassis, two Supervisor 6-Es can run redundantly, with Cisco claiming to have reduced the failover latency from 200ms to 10ms.

It's encouraging to see Cisco continuing to move away from the dark ages of PCMCIA flash cards and serial consoles, as well as upping the throughput in their midrange 4500 products. Even more encouraging, if the new supervisor for the 6500 series does what Cisco claims, it will mark a turning point in redundant networking.