RFP for a Premium Network, Take II

What's New at C2G

As noted, this year's RFP was a little different than the 2003 version (find last year's questionnaire and responses; this year's can be found later in this article). For example, last time we limited the field to vendors that could provide two 10-Gbps Ethernet ports; this time we required six 10-Gbps connections to campus buildings. We again asked for both 100-Mbps and 1,000-Mbps desktop links, but instead of also asking for PoE and non-PoE variations, this time both the 100-Mbps and gigabit desktop setups required PoE.

The designs from all three vendors were similar in that each provided a high-end switch/router at the core connected to another switch, or "aggregator," in each campus building basement. These aggregators fed the desktop switches on each floor. However, Alcatel and Nortel provided a single chassis in both the core and the basement, while Cisco offered dual switches in the core and in each basement of the 12 buildings. Expensive, but effective.

Each vendor took advantage of available fiber for redundant connections back to the core in the data center. This included the six gigabit-connected buildings and the six 10 Gigabit-connected buildings. Cisco provided redundant power supplies and fans in each of its chassis; Alcatel and Nortel added dual CPU cards and redundant switch fabrics, which helped compensate for their lack of totally redundant chassis. We didn't specify how much redundant equipment we wanted, but we did emphasize high availability. We could interpret the responses as an indication that Cisco felt it could provide five-nines reliability only with expensive, redundant chassis, and that Alcatel and Nortel each believed they could provide it with more redundancy within their switches. However, no vendor could provide any uptime guarantees. In most cases, a middle ground makes sense. At a minimum, C2G should have redundant switches at the core and spare components for everything in its network, including chassis. However, if you have redundant equipment set up in high-availability mode, you should have enough time to replace faulty hardware before the backup also goes bad--you don't need spares where you have "hot spares."