So this isn't a problem with latency then? My plan was to have only one opensips guest on each host. I just want the benefits of the vm guest but don't need to run anything else on the host so that everything can be given to that one guest. The machines would be dual CPU with plenty of memory. I'm not quite clear yet on what opensips requirements are for resources other than obviously, very good I/O.
Mike On Fri, 3 Jul 2009 10:25:51 +0100, Stanis³aw Pitucha wrote: > I'd go with: > > You can, but don't expect wonders. For example, check your maximal > possible throughput on udp - I've seen that some kernels can't cross > 1mbps boundary (just by blocking the queue, so it's the program > slowing down on send()). They were ok with tcp at the same time. > Also delays go up a bit too, so expect to see much more > retransmissions. You can start receiving ~20% more registers if you > check the database for example. > If you have very little traffic, guests have tendency to 'freeze' (at > least on local esx) when they don't do anything and then they take 3-4 > secs to process the first packet, but they're quick again for the next > one. > > So... sure it's possible, but if your esx hosts are already stuffed > with guests, I wouldn't recommend it. > > _______________________________________________ > Users mailing list > [email protected] > http://lists.opensips.org/cgi-bin/mailman/listinfo/users
_______________________________________________ Users mailing list [email protected] http://lists.opensips.org/cgi-bin/mailman/listinfo/users
