Gigabit Ethernet switches fast in benchmark tests
As gigabit ethernet becomes a cost-effective means of connecting high-performance servers to high-speed networks across todays ubiquitous Category 5 copper cable, Cisco Systems Inc. has finally made its first foray into this promising new arena.
Actually, Cisco has taken two steps into this market: IT managers of midsize sites will find the companys fixed-configuration Catalyst 3550-12T will serve well as a wiring closet aggregate switch, while Ciscos Catalyst 2950T-24 sibling is a decent choice for uplinking Fast Ethernet desktop segments to the Catalyst 3550-12T.
Both products were released last month. The 1.5U (2.4-inch-tall) Catalyst 3550-12T is priced at $9,995, or $833 per port. The 1U (1.75-inch) Catalyst 2950T-24 is $2,395, or $92 per port. By comparison, Ciscos Catalyst 3524XL, a 24-port 10/100M-bps switch with eight Gigabit Ethernet Fibre Channel ports, is priced at $90 per port. The 24-port, 10/100M-bps Cisco Catalyst 2900XL switch is $75 per port.
Although copper ports are much more expensive than Fibre ports, the copper wire that is used for Fast Ethernet is already in the wall at most companies.
Reaching the summit
overall, the catalyst 3550-12t aggregation switch performed well within expectations, although eWeek Labs performance tests produced slightly different results than did Ciscos lab tests (see story about the discrepancy, below).
The 12-port, 10/100/1,000M-bps device doesnt set any new standards for products in this category, which include the nearly year-old Summit5i switch from Extreme Networks Inc. (which is priced at $10,995, or $916 per port). Currently, the major justification for buying the Catalyst switches is the Cisco brand.
We put the Catalyst 3550-12T through its paces with a Spirent Communications Inc. SmartBits 6000B performance analysis system loaded with 12 Gigabit-over-copper ports. We also used Spirents AstII test software to run Layer 2 traffic tests.
We did not run performance tests on the Catalyst 2950T-24, which is a standard 24-port, 10/100M-bps fixed-speed switch with the addition of two Gigabit-over-copper ports that are intended for uplinks to the 3550.
Of course, placing too much emphasis on performance testing might be pre-emptory. All Gigabit-over-copper products are constrained by two main factors.
For IT managers, the most important of these is the fact that, despite the vendor hoopla, its not clear that many organizations need all this bandwidth, even for server farms: The server operating system and attendant applications are usually the performance bottleneck, not network bandwidth.
For example, applications running on Windows NT typically send between 2M bps and 4M bps of information to the networka far smaller workload than the 1,000M-bps capacity of Gigabit NICs.
Second, although prices for Gigabit- over-copper NICs are falling, these cards are still far more expensive than the cost of daisy-chaining several Fast Ethernet cards together and then balancing the traffic load among them.
How suite it is
Administrators will likely find it a little easier to monitor and configure these switches because of improvements Cisco has made to its Cisco Cluster Management Suite software, which comes with each product. We used the Web-based interface to monitor and configure ports on the Catalyst switches as well as on a couple of Cisco 2900XL switches we had on hand.
We found it easy to use the Catalyst 2950T-24 to integrate workstations and servers into the test network via the 24 fixed 10/100M-bps auto-sensing ports. Using one of the two Gigabit 1000T copper ports to uplink the Catalyst 2950T-24 to the Catalyst 3550-12T, it was simple to manage both devices from the same Web interface.
The Catalyst 2950T-24 comes with many features that wed expect with advanced managed switches. We were able to set up VLANs (virtual LANs) to manage broadcast and multicast traffic. Using VLANs also made it easier for us to manage security on the switch by limiting traffic between different switch segments.