3845 or 7600 for routing+switching design

[cisco_fcrochet]

Hi there,

I need to build a network architecture for 100 serveurs with 1 to 4 IP physical interfaces (10/100/1000 Eth.). Each interface belongs to different LAN/WAN/MAN, on each several VLAN have been created. The goal of this new design is to build a new switching/routing architecture. Question is simple: i've got to options 3845/3825 or 7600 routers ? But do i need to add switches before these routeurs as i'll need hundreds of ports ? How do i managed the connectivity between stacks of switches and the routeurs ? Thanks in advance for your info.

[Collin Clark]

Why not use a switch like a 6500 with layer 3 capabilities? Scales better and has better performance.

[lgijssel]

For a solution like this, you should rather look at a layer3 switched solution to interconnect the servers. It gives you more performance and flexibility. Depending on the amount of money that your company is willing to spend on this solution you may either use set of 3750's, a 4500 series switch or a 6500.

A router will only be required for the external connections. You can select the proper model based on criteria like throughput, required interfaces and communication protocols. Using a proper switching solution, you might be able to settle for a simpler router, i.e. 2800 or 3800 series.

Regards,

Leo

[albert.remo]

Hi,

Good Day!

For your query, if you're looking for resiliency and high-availability since you seem to be setting up a server farm, you can use a Cat6500-E multi-layer switch (capable of both routing and switching). C7600 is also a good option but quite high end if these servers would be deployed to different sites. Most modules and controller cards can be used to both C7600 and Cat6500.

Smaller routers would be beneficial for small nodes. You can use the 3800 series but you have to determine how many switch ports (module) you would require.

If these are critical servers, always put in mind high-availability and reliability. Look for equipment that are designed for these and as much as possible eliminate single points of failure. Have a separate switch to aggregate all your servers, then use a router to route traffic to other nodes.

Hope this helps.

Regards,

Albert