Quad T-1 per-packet sharing on 2620XM 1

[gnasses]

Hello,

Have a customer running 2 t-1 link on a 2620XM with per-packet load sharing and fair-queue, everything runs great. Customer wants to upgrade to quad (4) T-1 and they have 2 available WIC slots to add cards to. An engineer from the LEC says that they will not get good performance from their 2620XM with 4 t-1s and they should look at a 2800 series and an ATM DS3, which is quite a spendy option.

I am looking at the specs and think that 30K pps throughput should be anough to handle 4 T-1s with static routing, but I dont want them to buy extra WICs if they wont get a true 6MB pipe. This is a stub site and there are no routing protocols, no VPN tunnels, nothing real fancy. Looking for some ideas if others think this will perform adequately.

 

[BuckWeet]

You should definately upgrade to a new faster router.. If traffic ever needs to be shaped, QoS, etc... you'll run into problems..

 

[gnasses]

Yes, I am aware that they will need to upgrade for any additional services or demands, and we have made the customer aware of this. Eventually, they will be on a 2812 with DS3 over ATM, but this is probably 270-360 days out. DS3 + router cost would be a significant expense for them right now versus 4 t-1 lines + expense of new router. We just want to get them by in the meantime, or at least give them all their options if its possible.

My question is whether the 2620XM with 30k pps througput can handle throughput and per-packet bonding of 4 T-1s with static routing only, no QOS, no routing protocols, no VPN tunnels.

 

[BuckWeet]

You should talk to the carrier.. 4 T1's can be quite expensive, almost to the point where you could get a fractional DS3 for the same cost..

As for the 2620XM handling it.. Probably... Call Cisco, they could give you a better answer.

 

[JOAMON]

Question...
Are the current two T1 point to point links to the same end point??
If so then instead of having 2 1.5 links with per packet load sharing why not bond the two T1 circuits into a multinlink interface for one 3 meg pipe.
Dont see why that router could not handle data transfer over 4 t1 with no advanced features enabled. Check your CPU history at present with two t1's and if you have peaks above 50% percent then you could probably expect that with 4 you would have peaks over 100% (BAD). Keep in mind that the new 2811 router will use the newer wic-1dsu-t1-v2. So if you purchase additional t1 cards do not buy the older wic-1dsu-t1 as they will not function on the newer routers. The newer style wics will work with both the 2600XM and 2800 series.

 

[gnasses]

yes, all T-1 lines go to the same place. What is the difference between Bonding into multilink interface and per-packet load sharing, both concenptually and ios command line?

As far as I know CPU usage has been low for them, but we don't monitor or log it realtime.

 

[JOAMON]

A show process cpu history will show up to the last 72 hours.

Load balancing uses the two equal cost paths and balances the load across multiple paths. Your down/up speed will not exceed 1.5 even with two circuits. Bonding the T1 lines using ppp multilink you will create a new multilink interface and then add your T1 circuits into the multilink group. This way you will actually create one 3 meg pipe using the 2 T1 circuits.

Multilink might look like this:

interface Multilink##
description Maple Street
ip address x.x.x.x 255.255.255.252
no ip directed-broadcast
no cdp enable
ppp multilink
no ppp multilink fragmentation
multilink-group ##

interface Serial0/0
description
no ip address
no ip directed-broadcast
encapsulation ppp
no keepalive
tx-queue-limit 26
no fair-queue
no cdp enable
ppp multilink
multilink-group ##

interface Serial0/1
description
no ip address
no ip directed-broadcast
encapsulation ppp
no keepalive
tx-queue-limit 26
no fair-queue
no cdp enable
ppp multilink
multilink-group ##

The following command might be different on newer IOS
no ppp multilink fragmentation
Might read like
ppp multilink fragment disable

 

[BuckWeet]

how far is the remote end??

you need to worry about out-of-order packets..

say you have a long distance T1, and 1 T1 is 300 telco miles away, then the 2nd t1 is 800 telco miles away, you'll have some lovely problems..

I'd call to verify that the T1's are built identically through their network mileage wise..


BuckWeet