On Sun, Nov 4, 2012 at 3:36 PM, Alan Robertson <[email protected]> wrote: > On 11/03/2012 04:05 PM, Alan Robertson wrote: > * Memory footprint O(hosts log network-hops) due to over zealous ECMP, > > In order for me to preserve my ever-precarious illusion that I know what I'm > doing ;-), I looked up ECMP - Equal Cost Multi-Path Routing. > > One of the nice things about the topology-sensitive neighbor heartbeating is > that even for a very large number of hosts and a goodly number of sites, > very few packets cross sites - or even subnets. > > So, even if this were to somehow occur, and cause memory inflation, it > should only affect a small number of systems, because the overwhelming > number of systems would not be subject to routing at all (intra-switch > packets typically avoid routing).
Yeah, I was proposing the risk of a bug in the code that infers topology, such that it actually does cross switches, and wanted you to think about how you'd catch that in testing. > Of course, this assumes that I have implemented the code that I haven't yet > done, and that code does understand the topology, and that it is > sufficiently bug-free. For most networks, that should be by far the normal > case. > > Figuring out where subnets begin and end is typically straightforward. > > But as I should have noted before (and now have in the test plan), this > first release won't include more than one ring - which potentially brings > this back up -- except... > > I wouldn't think of this first release as an ideal candidate for deploying > over thousands of servers - so hopefully this should be avoided for this > release anyway. If you write up what the automated test should be, and point out that the testing isn't currently being done, then people can infer to what extent they're exposing themselves to a risk. > -- > Alan Robertson <[email protected]> - @OSSAlanR > > "Openness is the foundation and preservative of friendship... Let me claim > from you at all times your undisguised opinions." - William Wilberforce _______________________________________________ Assimilation mailing list - Discovery-Driven Monitoring [email protected] http://lists.community.tummy.com/cgi-bin/mailman/listinfo/assimilation http://assimmon.org/
