T1 aggregation limitations

Hello,

I want to use a pair of Cisco 2821 routers, each with 3 WICS (VWIC2-2MFT-T1/E1) connecting a total of 6 T1 lines to create a 9Mbs point-to-point link between two offices.

I want all lines to be load balanced so that any user could potentially use most of the bandwidth.

My understanding is that I can use CEF and enable ip load-sharing per-packet on each interface in order to accomplish this.

Can anyone confirm, based on experience, that this is a sound plan?

If this basic idea is sound, what options are available for setting priorties by MAC or Client IP address or otherwise limiting/controlling the burst rates?

Multilink PPP seems to load-balance based on IP address, and therefore, if my understanding is correct, no ip conversation would be capable of exceeding 1.5Mbs. Is this correct?

I appreciate any assistance you can offer.

Xophry

Reply to
Xophry K'un Le'im
Loading thread data ...

Multilink PPP is the way to go. One flow can use the entire pipe with MLPPP and the load balancing is almost perfect.

Got to ask the question though, for a 9 mbps link, have you looked into a MetroE or MPLS connection instead traditional T1s?

Reply to
Wayne

No it doesn't.

For N lines it divides _each_ packet into N pieces and sends one piece down each line.

Sucks up a lot of CPU. Maybe a 2821 will do though.

Would want some test numbers.

Reply to
anybody43

We use PPP Multilink in our network. We have 4 sites that use multilink bundles ranging from 5 to 7 T1s. We are using 2620 routesr on most of these links, and it does make the CPU work hard. During peak traffic times, the CPU goes to 75%, most of which is interrupt processing.

We have 1 2650XM that, under the same traffic loads, goes only to about

25%.
Reply to
Mark Williams

Ok, so Multilink PPP is not as limited as I thought.

Regarding traffic shaping/bandwidth allocation...

I have a mission crictical database application that will require 40% of the bandwidth if all current database users are bursting at the maximum rate simultaneously (rarely, if ever). Normal database traffic will average 15-20% of the total available bandwidth of the 6 T1 lines.

Additionally, I need to support normal resource sharing traffic. I want the database application to be able to utilize 40 percent of the bandwidth if it needs it, and I want the resource sharing/network management to utilize up to 95 percent of the bandwidth, if it is not in use by the database application. The mission critical use of the database can be isolated to certain IP addresses. Though other IP addresses will require occassional database access, these ancillary workstations do not necessarily require high-priority database access.

What is the best bandwidth management strategy for this? I was thinking that Class Based shaping would get me what I want, but this article does not indicate that this feature is available on the 2821 and I can't seem to find the correct documentation to answer my question.

formatting link

Thoughts?

Reply to
Xophry K'un Le'im

Hi, CBWFQ is actually avaliable on the ISR's (i.e2800).. and think its avaliable on the lowest avaliable feature set as well. Check the Cisco Tools page for the IOS feature navigator to be sure. It would work well for what your trying to achieve above. Ie. Reserving Bandwidth in the event of of planned bursty congestion. No good reason to use LLQ since little on no serialisation delay on a 9mb pipe.

Reply to
jay

Thanks for the direction.

Reply to
Xophry K'un Le'im

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.