2921 and BGP

[darren.g]

Gentlebeings.

Current $POE has projected that it wants to upgrade our internet links significantly int he next 6-8 months.

Currently, we have our own AS and allocated subnet, with two providers - so we're running BGP - but we're only using the second link for inbound traffic and backup outbound, as it's much smaller (primary link is 20 Mb/s, secondary is 2 Mb/s). Ergo, we only take a single default route from our "primary" BGP peer, and advertise our networks out both with some AS-prepending to affect the inbound traffic. All this on a 2921/K9 router.

We're looking at increasing to dual 15 Mb/s links *and* load balancing properly - which means effectively taking the full BGP table from *both* ISP's.

My question to you, learned ones, is will the 2921 take the load required to do this? I'm not so concerned with the throughput - the 2921 is rated to 75 Mb/s, which means 2 x 15 Mb/s at full duplex (60 Mb/s total) should be within its throughput limits - but I am worried about the BGP processing and routing table size.

I'm not sure how much RAM the 2921 has, but knowing the way things worka round here I'm sure it's the minimum specification (512 mb, according to the Cisco page located here - http://www.cisco.com/en/US/prod/collateral/routers/ps10537/data_sheet_c78_553896.html) - if I max out the RAM to the rated 2 Gb, will the RAM/Processor deal with running this configuration?

Input aprpeciated, and thanks.

Cheers.

[Joseph W. Doherty]

Disclaimer

The    Author of this posting offers the information contained within this    posting without consideration and with the reader's understanding that    there's no implied or expressed suitability or fitness for any  purpose.   Information provided is for informational purposes only and  should not   be construed as rendering professional advice of any kind.  Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In    no event shall Author be liable for any damages whatsoever  (including,   without limitation, damages for loss of use, data or  profit) arising  out  of the use or inability to use the posting's  information even if  Author  has been advised of the possibility of such  damage.

Posting

Internet BGP tables do take a toll.  The 2921 probably could deal with it with 512 MB, but you might try "forgetting" full BGP route tables beyond taking a default, just run OER/PfR.  Besides dynamically load balancing, it finds best path per destination based on actual performance and actually can use much less CPU.

[darren.g]

JosephDoherty wrote:
 
Internet BGP tables do take a toll.  The 2921 probably could deal with it with 512 MB, but you might try "forgetting" full BGP route tables beyond taking a default, just run OER/PfR.  Besides dynamically load balancing, it finds best path per destination based on actual performance and actually can use much less CPU.

Interesting suggestion, but not an area I've had any experience with, so might be a bit of a stretch to configure from scratch.

Who would you take a default from? One ISP? Both? And how do you pick - flip a coin?

Still, I'll bear it in mind as an option.

Thanks

[Joseph W. Doherty · Accepted solution]

Disclaimer

The     Author of this posting offers the information contained within this     posting without consideration and with the reader's understanding  that    there's no implied or expressed suitability or fitness for any   purpose.   Information provided is for informational purposes only and   should not   be construed as rendering professional advice of any kind.   Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In     no event shall Author be liable for any damages whatsoever   (including,   without limitation, damages for loss of use, data or   profit) arising  out  of the use or inability to use the posting's   information even if  Author  has been advised of the possibility of  such  damage.

Posting

You would either take a default from both ISPs, or you can use a static default to both ISPs.

What happens, outbound traffic will initially round robin between both ISPs, but then OER/PfR will adjust, if necessary, load on each link, to balance them, and/or move individual flows if it sees one path performing much better than the other for that destination.

I realize at first look OER/PfR looks complex.  And, it can be with all its options (especially PfR).  But a configuration just supporting dynamic load balancing and passive flow analysis isn't really too complex.

Years ago, I dealt with a pair of 3660s each with a DS3 connection to an ISP, each with both ISP's full Internet route tables.  The DS3s would run to capacity and the CPUs on the 3660s were running at capacity too.  Changed them to OER, using static defaults, and CPU load halved.

When you think about it, you should be able to reach any destination from either ISP.  Having full Internet BGP tables allows selection of "best path" to any destination, but what really is "best path"?  It's usually least AS hop counts; which may have nothing to do with physical hops, actual link bandwidth, and especially actual available bandwidths.

OER/PfR can analyze end-to-end performance.  So, for instance, OER/PfR could determine sending traffic to your site via your 2 Mbps link, at some times, might provide better performance than your 20 Mbps; which certainly could be true if your 20 Mbps link was running at capacity and the 2 Mbps link idle.  (NB: this doesn't mean to make your two links equal, because today, remote sites are not running such analysis; although today OER/PfR could proportionally load balance across both your 20 and 2 Mbps links, outbound.)

[darren.g]

Joseph.

Thanks for that - I'll certasinly bear it in mind (and may even look at implementing it NOW instead of waiting until the new ISP links are approved).

At least I can prove the current hardware will deal with what is necessary before we get to the "What do you need for the new links" stage.

Cheers