Cascading Switches Vs Chassis Based

[mnlatif]

Hi,

We currently have 32 Fiber Ports (4 3508XLs) and 96 10/100 (2 3550s) in our data center. We will be needing more Fiber\Copper capacity in coming months. I am trying to evaluate between adding more individual switches OR Replacing the existing setup with a Chassis Base solution (E.g. 6506\6509). My question is the Justification for this with concerns

1. Isn't a chassis based solution more suspectible to service disruption ? As the SUP module failure can bring the whole system down as compared to single-switch failure, which only effects the specific ports.

2. What ADV does a chassis base system provides, if we only intend to use Fiber\Copper Ports ? No IDS\FW modules will be installed.

3. Is there any cost advantage ? As apparently it seems that a 48-port blade is almost the same price as an individual switch.

4. Any performance advantage ?

Any comments\personal expriences ?

Regards \\ Naman

[mhussein]

The main advantage of a 6500 over a cluster is switching fabric bandwidth. In a cluster you have multiple switches connected to each other via one or more gigabit port ether-channels, that gives you 5Gbps max bandwidth between each switch hop. Compare that to the 32Gbps backplane of the 6500 with no ports wasted on ether-channels.

6500 with Sup2-MSFC gives you up to 210 Mpps layer 3 forwarding, according to Cisco's specs, versus < 7Mpps on a 3508 connected to a router. I think these numbers affect metrics such as latency and jitter as well.

Other things to consider is the high availability of the 6500 (redudnant Sup, power) versus multiple points of failure in a cluster.

Cost wise, you have to pay a steep price if your network is expanding into an enterprise or service provider level of performance and scalability requirements. That goes for most of the "extreme" high-end switch manufacturers ;)

One more thing to consider is the scalability and the multitude of 6500 switch modules.

That is my 2-cents, I hope others would share thier experience on how they perform network capacity sizing and planning, and how they make thier case for major investments on network equipments.

Regards,

Mustafa

[amitsin]

Naman,

All the major and imp points are already highligthed by mustafa and just to add one more :

on 6K's with supervisor engine along with MSFC and PFC gives you hardware l2/l3 switching + Cef and TCAM acrhitecture in which ACL's look up are done in hardware which gives better performance with no/less CPU load.

HTH

AS

[pswale]

Hello,

We also recently faced the same sort of situation. Another thing you can consider is a stack of 3750 switches which have a backplane of 32Gbps. They are cheaper but they don't have Power redundancy unless you purchase the RPS as an option which plugs in and provides DC power to the switch if AC fails.

I personaly think the Chassis solution is superior because the MTBF is 10X that of a access switch ie 2950, 3500, 3550, 3750. The reasons for a chassis also go into techincal advantages that are geared more for performance and optimization for a busy network. If your current network utilization is not significant now, you need to consider the future and where it plans to grow.

thanks,

swale