Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations Chris Miller on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Cabling & Design of Network

Status
Not open for further replies.

skk391

Technical User
Mar 3, 2009
332
GB
I have a question regarding cabling a network. I have taken over a network. I have ran cdp commands and mapped the network & I am not impressed. The network consists of around 8 * 2950 switches, 2 * 3500 & 2* 2960 gigabit switches. I have 20-25 server and about 200 end users.
It looks like the old network admin just connected crossover cables within the switches in any randon order and hoped for the best. I can see that if the cable went bad within two switches I would lose half of my network straight away.

I would like to build redundancy into the network. I understand the concept of STP and how it builds redundancy into the network. But I am unsure on how to cable it all up. I am even thinking of configuring Etherchannels and take advantage of load balancing.

If I was take the concept of core, distribution and access, would the best design be to connect all servers to the 2 * gigabit switches and then connect each of the other access switches to the gigabit as well, if one of the 2950 switches was to fail or the trunk was to go bad I would only lose users on that switch rather than half my network which is what would happen at the moment, because it seems to just be daisy chained switches.

If I do connect all of my switches to the gigabit switch the only problem I see is that I will be using a lot of ports on my gigabit and I might not have enough ports to connect my servers. i.e

2------2950
9------2950
6------2950
0------2950 etc

If I don’t connect all of my switches to the gigabit switch. Would it be a good idea to maybe daisy chain no more than 2 switches of the gigabit switch and configure etherchannels and build in redundancy that way i.e.

etherchannel configured here
2960--------------2950--------------------------------2950
--------------2950--------------------------------2950

I just want to get away from a single point of failure, having said that with the design that I have mentioned above, if the 2960 was to fail then the whole network fail..

Any help would be very useful. I think I am ok with the commands, I am more interested in the design of the network please.

Many Thanks


 
where is the L3 routing done and where does it connect currently ? If you have a single router doing layer 3 then you still have a single point of failure anyway.
 
Hello,
I have L3 routing but it is non cisco. I am just concerned about the L2 side at the moment and what the best method would be to cable this sort of network. Like I say all of the switches are just daisy-chained at the moment and I need some advise cabling.



 
I guess one option is to duplicate your existing links and etherchannel the two links between each pair of switches together.
Avoid relying on spanning tree if you can as link aggregation gives you better performance in the case of a link failure.
 
Ok thanks, how should this be cabled? Should all access switches be connected to the newer 2960 core switch. Or do you think that I should daisy chain a couple of switches of the 2960?

 
If at all possible, I would run each one back with a separate connection to the core switch.
A little more work/expense up front, but well worth it in the long run.
 
thanks for you input, what are the advantages of running each access switch back to the core switch?
 
Each link going back to the core from each access switch is so that if 1 link is cut, all of your other switches do not go down. I do prefer the etherchannel idea though, you have 2 links going from core to the access level and if one cable gets cut, you still have that one leg up and running.

------------------------------------
Dallas, Texas
Telecommunications Tech
CCVP, CCNA, Net+

CCNP in the works
 
Running each access switch back to the core switch gives you a "star" topology.

It is generally accepted that a "star" topology is good practice in network design for reasons of resilience, performance and management.

Sometimes it isn't possible, and so you have to daisy-chain switches. Fine, but just don't choose that as your first option.
 
Thanks for the reply guys, Im going to take your advise and setup a start topology with etherchannel for resilience.

Thanks again :)
 
If your core switch needs to take a lot of fibre uplinks, something like a 3750-12 is very useful.
If you need copper ports in the core as well, you can stack the 3750-12 with a 3750-24 or -48.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top