2

Ok so I am taking over a network and I'm trying to figure out the best topology for it and looking for some advice on what you experts think.

Right now, the ISP demarc goes to an unmanaged switch (Switch 1). Switch 1 then has a link to the WAN port on Router 1 (Sonicwall 220), and 2 ports to the same Windows server (handling AD and DNS). DHCP is handled by the router currently.

Router 1 then has a link from it's WAN port to port 49 of Switch 2 (Netgear FS750T2). Then the rest of the managed devices are as follows:

Port 50 Switch 2 -> Port 49 Switch 3 Port 50 Switch 3 -> Port 49 Switch 4 Port 50 Switch 4 -> Port 49 Switch 5 Port 50 Switch 5 -> Port 49 Switch 6

One if not two of those switches will be taken out, as only a few ports on each are really being used.

I have a couple questions:

  • Is going from the ISP -> unmanaged switch -> router OK? Or should the ISP go directly to the router and leave the unmanaged switch out? The Router will be changed to a Sonicwall 2600, giving me more ports on the router.

  • Should the switches be trunked on those ports at all? Is just a direct link OK? Or should each switch go to a port on the router? These are not VLAN'd at all.

  • I'm thinking the server should have 1 nic to different devices for redundancy. Right now both NICs are just set to .10 and .11 on the same network. However, where does that fit in if I skip the unmanaged switch (Switch 1) and going directly to the router?

Any help is appreciated. It's for a call center and I have a few weeks before things start getting crazy. So if I'm going to start taking things down and rearranging devices now is the time. Thank you for all your help ahead of time.

  • you should bring network topology instead of long description – Gadeliow Sep 30 '15 at 23:36
  • Is the windows server providing any type of external service? Like VPN or RAS gateway? – cpt_fink Oct 1 '15 at 7:09
2

enter image description here

it seams like you have network like shown
first of all it is not proper network design , where cascading switch connection means single point of failure and traffic bottleneck.so you should run star topology.
second wan port refer to external connection ,so your router not connected to SW2 via wan port it just Ethernet port and in your case is internal

regarding your first question : no problem of have switch here for two purposes
a) may be in the future you get more services from this ISP (WAN , VPN ..etc) so each one may need single port
b) port in a switch is cheaper than port in router

regarding your second question : for any switch connect any other switch , ports involved in this connection should be configured as a trunk

regarding your third question : keep both link from your server to the switch and make use of great feature port aggregation or NIC teamming from the side of server , which will increase the Bandwidth and add redundancy to your server

sure for all unused ports on any switch you should shut it down or remove those switches cause it is consider one of security issue

| improve this answer | |
  • Thanks! and thanks for that diagram. Could I essentially trunk the switches together on port 49, then on port 50 connect them to the router? – DanielF Oct 1 '15 at 0:16
  • for the time being you may chose sw 2 to acting as a aggregated core and connect all switches on it (trunk for sure) and connect the router also on it (access no trunk) . and in the future try to get L3 switch with suitable back plane to acting as your core layer , network should be in stare topology not cascading – Gadeliow Oct 1 '15 at 0:22
  • Is there a huge difference in going directly to the router with the switches vs aggregating them on switch 2? Wouldn't switch 2 become bottlenecked? – DanielF Oct 1 '15 at 0:32
  • no there is no hug difference you just configure ports to switches as a trunk and port connected to router as a access . in the cascaded design port 50 in SW2 will process the traffic of the 4 downstream switches but in the stare topology each port will handle traffic of one down stream switch (so that i recommend you to get stronger hardware to acting as aggregated core ). and by the way in stare topology there is no single point of fail – Gadeliow Oct 1 '15 at 1:05
0

The SonicWALL is a firewall. It is best to have everything behind the firewall that you possibly can. You need to discover the reason that the server is not behind the firewall. It may be a legitimate reason, and you may be able to get around that reason.

Switch-to-switch links should be trunks. Call centers are usually sensitive to single-points-of-failure, and this network design has a bunch of them. Having the switches daisy-chained, especially such a large chain, leaves you vulnerable to having the loss of a single switch disabling most or all of the LAN.

In the current switch configuration, having the server connect to two different devices doesn't mean much. If the first switch in the chain quits working (for whatever reason), all the switches quit being able to get to the firewall.

| improve this answer | |
  • This was exactly what I was looking for thank you. So on trunking the switches, should I have a trunk from Sw1 -> Sw2, Sw2->3, and Sw1->Sw3? Just trying to figure out the single point of failure like you mentioned. – DanielF Sep 30 '15 at 23:23
  • Yes, you should trunk each link from one switch to another. If the first switch in the chain fails, none of the users on the other switches can access anything beyond that switch (server, Internet, etc.). The single firewall, server, server switch, Internet connection, etc. are all single-points-of-failure. You may not be able to do anything about most of them, but daisy-chaining switches leaves hosts on any switches after a failed switch off the network. That is something you may be able to correct. – Ron Maupin Sep 30 '15 at 23:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.