Oversubscripted Bandwidth Points on 7206VXR

[iwanc]

Hi,

I just want to ask about bandwidth points on cisco 7206VXR with NPE-G1.

What will happen if bandwidth points oversubscripted on one of the PCI bus?

This is the sample:

===========

PCI bus mb1 (Slots 1, 3 and 5) has a capacity of 600 bandwidth points.

Current configuration on bus mb1 has a total of 0 bandwidth points.

This configuration is within the PCI bus capacity and is supported.

PCI bus mb2 (Slots 2, 4 and 6) has a capacity of 600 bandwidth points.

Current configuration on bus mb2 has a total of 640 bandwidth points.

The set of PA-2FE, PA-POS-2OC3, and I/O-2FE qualify for "half bandwidth points" consideration, when full bandwidth point counting results in oversubscription, under the condition that only one of the two ports is used. With this adjustment, current configuration on bus mb2 has a total of 340 bandwidth points.

This configuration is within the PCI bus capacity and is supported under the above condition.

==========

What happen when 2 ports of PA-POS-2OC3 activated?

Thx

Iwan

[paolo bevilacqua]

It happens that due to oversubscription of the internal bus, as soon there is a little of bursty traffic, packets will be mysteriously dropped, other strange things will happen, cisco TAC will NOT support the router, and you cannot utilize your expensive OC3 circuits.

That is why you never want to exceed the "points" limitations of the 7200 routers.

[iwanc]

Is there document related to your answer? I cannot find it.

thx

iwan

[paolo bevilacqua]

I think that when cisco says configuration exceeding bandwidth point are not supported, that is enough to make the point clear.

It is not that they must document whatever behavior an unsupported configuration may cause.

[iwanc]

Yeah I know that. But I need more specific document for my customer to convince them not to put modules with Oversubscripted bandwidth points

[bcoverstone]

I know this post has been sitting idle for 6 years, but I wanted to add something.

The oversubscription thing can actually be a bit overrated, depending on your situation.  What Cisco has done is take the WORST case scenario, and applied equivalent points for each module.  That way, their TAC is guaranteed that the issue will not be related to oversubscribing the bus.  From a tech support point of view, and for their sanity, I can understand that.

Now, for the real world.  Most of your cards may only be utilizing 40%, 30%, some even 10% of their capacity.  For example, I have a Cisco that has 1400 points on one side of it.  Reason being, I have a PAS-VAM2+ card in there that takes 600 points, a 2 port FastEthernet card, and then the built in 2 port fast ethernet I/O controller.

In reality, I replaced an older PAS-VAM with the PAS-VAM2+.  We only ever encrypt a max of 10Mbits of bandwidth (the old PAS-VAM was rated for 150, the new PAS-VAM2+, I don't even know how high it goes).  We keep three of the four fast ethernet interfaces administratively down.  And the final FE port only uses about 30-40% at any given time.

So according to Cisco, we are using 1400 points.  But in reality, we are using about 100-200 points.  Do we ever have a problem?  No.

You have to use some common sense here.

[hennigan]

Is there a reason not to move one of the PAs to an odd-numbered slot?

[iwanc]

Yeah, there are some users connected to the existing PAs. They are concern about users downtime.

[hennigan]

72xx PAs are hot-swappable. What we've done in similar situations is to schedule a maintenance window, copy the configuration to a text editor and search/replace the affected interface numbers, add a "no shutdown" on the config to the physical interface to which the card is going.

Console in. At the scheduled hour yank, shove, paste the config. Your customers on the affected card should be back up in less than two minutes.

Not really relevant to this specific issue but it is generally a good idea to have a specific maintenance window at specified off hours for singly-homed customers (and even multiply-homed just in case). 2:00 AM Sunday morning local time as an example. Situations like the above, upgrading IOS for security fixes or features, etc. will always come up from time to time. If it's expected it usually isn't a big deal and they often appreciate the attention to keeping your network up to date.

Back to your original question, things will still work in an oversubscribed state unless you're pushing near full wire speed in which case you'll start to get packet loss. If you need to open a TAC case or get support you'll be told not to do it, and perhaps to fix the issue before getting further support.

This document:

http://www.cisco.com/en/US/products/hw/routers/ps341/products_tech_note09186a00800c814a.shtml

says "You can use a 7200 series router with a PA configuration that exceeds the guidelines that the Cisco 7200 Series Port Adapter Installation Requirements provide. However, in order to prevent the occurrence of anomalies while the router is in use, restrict the type of PA that is installed in the router in accordance with the guidelines. Additionally, in order to have Cisco Technical Support troubleshoot these anomalies, your PA configuration must be within the guidelines."

which may give you the ammunition you need.