Reasons to limit 1Gb capable switches to 100Mb?

tgrogan798 · ‎07-09-2012

Our network administrator restricts all our new 1Gb switches to only 100Mb, to which he claims it will increase overall performance. Preventing devices from bottlenecking the network. (that is, only restricting the main switch ports, not the uplinks)

For example we have a new building with 2x WS-C2960S-48TS-L connected together with FlexJack and then a 1Gb Fiber connection back to our core switch.

This building is on it's own subnet and there is little broadcast traffic. I don't see the point other than it hinders the potential speed we could use. Labs are set up in this building and 1Gb is MUCH faster when it comes to imaging and software deployment.

I've searched and can't find any documentation about limiting a switch for overall performance reasons.

I'm going to try to return them back to 1Gb but wanted to be sure I'm not missing something here.

Reza Sharifi · ‎07-09-2012

I agree with you. There is no need to limit bandwidth on user ports unless you have some sort of congestion issue on your uplinks. Use Gig speed and monitor it to make sure there is no issue.

HTH

mikegrous · ‎07-09-2012

This is ridiculous. So transferring a file will be 10 times slower all the time. verses putting everyone at gig and having everyone faster all time time, and slow only when there is a bottleneck. If he is truly concerned about bottlenecks and traffic QOS should be implimented. But only if there is congestion. Do you uplinks even get saturated?

patrick.preuss · ‎07-09-2012

Hi

Only reasons i know is your l2 (cabeling can not support 1g) infrastructure or your Servers.

On The other Hand it is maybe politic or policy.

Sent from Cisco Technical Support iPhone App

darren.g · ‎07-09-2012

Travis Grogan wrote:
Our network administrator restricts all our new 1Gb switches to only 100Mb, to which he claims it will increase overall performance. Preventing devices from bottlenecking the network. (that is, only restricting the main switch ports, not the uplinks)
For example we have a new building with 2x WS-C2960S-48TS-L connected together with FlexJack and then a 1Gb Fiber connection back to our core switch.
This building is on it's own subnet and there is little broadcast traffic. I don't see the point other than it hinders the potential speed we could use. Labs are set up in this building and 1Gb is MUCH faster when it comes to imaging and software deployment.
I've searched and can't find any documentation about limiting a switch for overall performance reasons.
I'm going to try to return them back to 1Gb but wanted to be sure I'm not missing something here.

I can see only two reasons.

1) The local cabling might not support 1 gig connections reliably - especially if it's old.

2) The uplinks might not support the aggregate total of the ports when being fully utilised.

Now, given that you've specifically said that this site is on its own subnet, and that there is little extra-site traffic, the second reason is not valid - UNLESS your local site is only layer 2 (I.E. doesn't have its own SVI or router interface for the subnet), in which case all routing decisions need to go back to your core - but even that would be unlikely to flood a 1 gig uplink, even if all 48 ports were trying to get routing decisions at the same time.

Maybe your guy is planning for a worst case scenario - but for mine it's a waste of time, and is only inhibiting the functionality of the site.

Cheers.

tgrogan798 · ‎07-11-2012

We're well over subnetted if you ask me. I was told the reason was because of some old computer monitoring software that was broadcast only and is something that is non-existant now. This subnet was given it's own block of 2048 addresses and a max of about 150 nodes (including wireless) will ever be used. (building is too small to fit anymore)

Routing decisions do have to get to the core switch, but that switch is only 150ft away in another building.

I'm about ready to try to take the keys away from him on handling this. He refused to move from RIP to EIGRP because it was only "for a large environment that needs backup routes and load balancing". Just because you don't need luxury doesn't mean you should drive a Ford Focus instead of a BMW when it's free.

mikegrous · ‎07-11-2012

He sounds lazy. You could convert to EIGRP over rip seamlessly. But if RIP is working there really is no reason to change. Of corse it all depends on your design etc.

tgrogan798 · ‎07-11-2012

Yeah, I think lazy when I hear excuses like that. The internet is frequently going down, but not for more than a few min at a time. I don't know when it happens each time, unless I personally can't connect or someone calls me and is asking why...

Of course, I call the guy and he says "I don't see any problems".

But judging by knowing it goes off at least a few times a day for a couple minutes I would guess it's RIP.

Anyone who's done it before knows that EIGRP is just as easy to configure in it's basic state as it is to enable rip. Plus we could use summary routes and significantly shorten routing tables.

Although I am CCNA, I don't have access to the routers or switches yet because I'm in a totally different position. If I did I would know for sure, or what else is causing it

nkarthikeyan · ‎07-11-2012

There might be other reasons..... Even though if we have 1g capable switch... our UTP endconnecting cable should support 1 G. If that backbone utp cable connects to end user is Cat 5 or some old cable then it wont supports 1g & also your laptop/system net card should support 1g.... We have faced similar problems in many cases where we had 1g configured..... You can request your network adminstrator to make that as auto negotiate for speed and duplex. So that if there is a feasible way to have 1g... it will operate with 1g... if not this will negotiate with 100 mbps and start working... hope this helps...