cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
498
Views
0
Helpful
3
Replies

T1 Service Degraded at 70% Utilization

JASON BOYERS
Level 5
Level 5

On Cisco's web site at http://www.cisco.com/univercd/cc/td/doc/cisintwk/itg_v1/tr1915.htm, they state:

"Excessively high bandwidth utilization greater than 70 percent results in reduced overall performance and can cause intermittent failures..."

I cannot find a good explanation as to why this is. Since a T1 is a full duplex circuit, why can't it go up to almost 100% utilization without degradation? I have seen the degraded service, but why is this? Thanks for the assistance.

3 Replies 3

jwelby
Cisco Employee
Cisco Employee

j.boyers: I am not sure what they are trying to convey. Unless at 70%, there is a queuing or buffering issue on the router. Or, are they saying you are not accounting for protocol overhead?

I believe that it is 70% utilization at the physical layer. That's based on slowness that I have seen on T-1s where they were being used at or above that level.

faheyd
Level 1
Level 1

I've read that whole doc a few times now. The best I can come up with is that while a T1 can transmit/receive at 100 percent with no problems, it might not be sending what your application needs at that particular time. You'll see that some of the section refers to 'bridging', and DECnet. While bridging across a serial link is not the best way to do things, I think this part of the document was trying to show what can go wrong with the serial link if you oversubscribe the available bandwidth. So, if you're doing file transfers and the link is 50 percent utilization and everything is fine, but when the link is near 90 percent and you get failures, the document was trying to give you possible solutions and indications of what the problem is. In the old days with Novell logins over a serial link, the 'login' app had no error control built-in. If the serial link was oversubscribed, the link would drop packets. This would cause the login to fail. The link looked fine, pushing all it could, but we could see buffer overruns and other things. The way to fix it was to 'prioritize' the IPX traffic.

I hope this helps.