[lvs-users] How can I improve LVS performance?
paulo.s.rodrigues at gmail.com
Wed Aug 5 23:49:28 BST 2009
On Wed, Aug 5, 2009 at 7:07 PM, Graeme Fowler <graeme at graemef.net> wrote:
> On Wed, 2009-08-05 at 10:10 +0100, Paulo Rodrigues wrote:
> > I've tried a few hosts, all giving pretty much the same results, in the
> > subnet (18.104.22.168), in another subnet (22.214.171.124) and in a different
> > subnet but on the same switch (126.96.36.199), both on gigabit and 100Mb.
> For consistency, can you please do your tests again from a single test
> client host, preferably on a different network to the director and the
> realservers, but one which can reach both the director and realservers?
> Also, when you connect to the realservers directly are you absolutely
> certain that you're getting the same content delivered as when you go
> through the LVS?
> Please read the documentation before posting - it's available at:
> LinuxVirtualServer.org mailing list - lvs-users at LinuxVirtualServer.org
> Send requests to lvs-users-request at LinuxVirtualServer.org
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
I'll have to change the network around to be able to do this, perhaps over
In the meanwhile, your words made me think of something else, I was certain
the delivered content was the same in every case, but the page is somewhat
large, around 110KB, so I repeated the tests with a tiny html page (80
bytes) and a tiny php (same 80 bytes but also connecting to a MySQL DB)
page; the same setup that serves 6000 large pages per *minute*, now serves
20000 tiny html pages per *second*, or 4500 tiny php pages per *second*.
The limiting factor seems to be, after all, the actual bandwidth. Part of
the network plumbing that I took for all-gigabit must be 100mbps.
LVS rocks, and I failed :)
Thanks for your help!
More information about the lvs-users