Network design

Hi,

Would somebody please be so kind as to point me to some information concerning network design for a data center? I've not had much luck looking myself, probably due to ignorance. I'm looking for stuff that's written with experience behind it. I want to compare my ideas with those of somebody who's done this for a while. I'm also looking for equipment suggestions.

Basically (I haven't drawn this out yet, I apologize) I'm looking at starting small and building from there. So this is how I envision things at this beginning point. We're only going to have one internet connection to start (our customers are more concerned with data security then availability, and we can't afford multiple lines until we have more customers!)

So, from all I've read so far, we'll need one router sitting on the "border", with the 'net connection run into it. When we move to a redundant network, the other internet feed will run into a seperate "border" router. From the router, the traffic will pass through a firewall. From that firewall, it will go to an interior router (core router?). From that core router, it will get distributed to switches, which distribute it out to clients (who might have their own switches, etc.)

Is this generally the way things are done? We've already got redundant power (UPSs/generators), so that's taken care of. Now we're just trying to design the network. Keeping in mind we will be starting small (somewhere from what we've got not (1.5mbit) to 10mbit or so) for at least the next few months, what kind of equipment would you suggest? A different design? Am I making any fundamental errors? I realize lacking internet redundancy is one issue, but we can't do anything about that until we have more income, and right now the clients who we're in talks with don't care as much about connection reliability as they do data security. There is a possibility we'll need to expand extremely quickly in bandwidth, but I suppose with money comes solutions.

Am I over-engineering things at this point? Is having a router sitting at the edge, passing through a firewall into an internal router, into a switch, going to the customer switch overkill for 10mbit? Or is it wise to design like this to future-proof things as much as possible? I'd like to be able to go up to 100mbit on the setup, but cost is most def. an issue at this point, we don't have money to burn, building a data center underground inside a mountain is costly. Thanks.

David J. Orman
ormandj (AT) corenode.com

[hostmedic]

For most of Cisco's site you will need a username and password - the good news is - this is free for a large portion of their site - but unless your a CCNA or above - some parts you cannot gain entry into.

I would first start with: http://www.cisco.com/en/US/netsol/ns...nces_list.html


There are some books in print as well - such as "Data Center Fundamentals - Cicso Press I believe: http://www.amazon.com/gp/product/158...89760?n=283155

I understand in a way why you choose not to give redundant paths - however I think you may be shooting yourself in the foot.

One suggestion may be to locate your facility (if possible) close to a larger network access point and purchase an additional backhaul if needed...

If you need more - let us know.

[SuperRacks]

To be honest, for 10Mbit I wouldn't care to split core and border. You could use a firewall with L3 capacity to construct a more secure internal network. With only one uplink, you wouldn't be short of interfaces either.

Instead, buy a second 'core' router and add some sedundancy. If the router burns down (for whatever reason), you might be waiting some time for replacement.... That will be enough to 'free' all of your customers to a competitor. Clients say that safety is more important, untill their server is down for 24 hrs. And a second uplink (to the same carrier with a different path, or to another carrier) really isn't that expensive for a couple of Mbits.