[c-nsp] IP SLA measurement against USA

[Anton Kapela]

On Jul 15, 2010, at 6:02 AM, LM wrote:

> Hi,
> 
> I want to control the SLA gave by our internet provider against USA and Europe.
> The issue is that I would like to configure SLA code in our border routers to have some visibility of it.
> So, is there anyone in the mailing list who can give me a fixed points -IPs- to verify this?, I don't know if could find a website with some destinations as a reference to configure my routers.

Without doing your own collocation of IOS boxen in the US to use as 'sla test points' it's unlikely that you'll have more than icmp echo at your disposal. Additionally, if one did have collocated routers to bang on, IOS RTR/SLA-based throughput testing will not scale anywhere close to gige, nor 10 gige. If your goal is to know about 'headroom' on circuits, and if you're operating at sufficient link rates, this method will only reveal the worst-case congested states in the provider upstreams (i.e. when DDoS is flowing over the same network/isp, etc).

If you're looking for well-known IP's to ping...don't. Using active probes to check SLA is an exercise in futility. With the advent of high speed ports, short queues, and the uptake and deployment of router CoPP over the years, active-probe based measurements will almost assuredly be 'wrong.' Worse still, they will be misleading, and will not expose anything of use to you.

This has even been studied in detail, and I recommend reading http://www.cs.wisc.edu/~pb/intcomp_final.pdf before you continue to explore this area. The 'take home' is, you will certainly not be detecting the real/actual loss, and if you do detect any, it will most certainly be an incorrect rate.

-Tk