We have a /20 of public IPs for our web, email, and other servers. I'm curious as to which is the better accepted practice... Should I keep all of the IPs together in one monolithic network or should I break it down into smaller segments using VLANs to logically separate traffic?
Right now (I didn't set this up) they're set up as multiple subnets existing all in the same layer 2 logical network and performance sucks. The router port is a 10mb port assigned with secondary IP addresses. Needless to say, I need to change this.
I have two 100mb ports available, so I was going to create a port channel and use dot1q trunking to separate each group of servers (web servers, email servers, phone servers, other appliances) into their own VLANs of smaller subnets.
Benefits I can see with this method is greater throughput and faster server-to-server communication (although this is generally rare). It will also give me greater control, and I'll be able to do things like restrict off a management subnet I can use to plug all my iLO ports into. Basically, I want more control over the network, and I think this is the way to do it.
But, is this way the standard practice or is it better to use a single gigantic subnet?
Right now (I didn't set this up) they're set up as multiple subnets existing all in the same layer 2 logical network and performance sucks. The router port is a 10mb port assigned with secondary IP addresses. Needless to say, I need to change this.
I have two 100mb ports available, so I was going to create a port channel and use dot1q trunking to separate each group of servers (web servers, email servers, phone servers, other appliances) into their own VLANs of smaller subnets.
Benefits I can see with this method is greater throughput and faster server-to-server communication (although this is generally rare). It will also give me greater control, and I'll be able to do things like restrict off a management subnet I can use to plug all my iLO ports into. Basically, I want more control over the network, and I think this is the way to do it.
But, is this way the standard practice or is it better to use a single gigantic subnet?
