|
> Right now I'd be more interested to find out why REDIRECT is 3 times
> slower than having the VIP on a non-arping device, to see if it can be
> fixed.
Logically, it's just a function of the way the redirect code operates.
I'm no kernel wiz (which means take what I say with a grain of salt), but
the way I am envisioning things here is as follows:
Without redirect:
Ethernet -> TCP/IP -> Application -> TCP/IP -> Ethernet
With redirect:
Ethernet -> TCP/IP -> Firewall/Redirect Code -> TCP/IP -> Application ->
TCP/IP -> Ethernet
Note that the second TCP/IP may or may not exist, it's merely speculation
on my part, however, it does make some kind of sense.
So, at the very least, you're going through the extra layer of the
firewall/redirect code, and imo, you're probably going through the TCP/IP
stack a second time -- the packet has to be reconstructed, and probably
re-processed by the tcp/ip stack.
That would definatly explain the slowdown, since _every single packet_
received is going to go through these extra steps.
The REDIRECT method is elegant and simple, but apparently, not without
cost.
Thanks,
Kyle Sparger - Senior System Administrator
Dialtone Internet - Extremely Fast Web Systems
(954) 581-0097 - Voice (954) 581-7629 - Fax
ksparger@xxxxxxxxxxxxxxxxxxxx
http://www.dialtoneinternet.net
On Tue, 15 Feb 2000, Joseph Mack wrote:
> On 15 Feb 2000, Doug Bagley wrote:
>
> > Looking at the source, it looks like
> > each service (VIP) has a list of real servers (destinations), and the
> > active connection count is by each RIP in a VIP, not by RIP alone.
> > That is, the active connections from one VIP are not counted in the
> > active connections for any other VIP.
>
> yes.
>
> I don't know the code real well..., but
>
> there is a hash table for each VIP and scheduling is done from it.
>
>
> > It seems to me it would be useful in some cases to use the total number
> > of connections to a real server in the load balancing calculation, in
> > the case where the real server participates in servicing a number of
> > different VIPs.
>
> there's lots of things that could be done and people fix what
> they think is worth their time to fix. This one could be a mess
> to change.
>
> Right now I'd be more interested to find out why REDIRECT is 3 times
> slower than having the VIP on a non-arping device, to see if it can be
> fixed.
>
> Joe
> --
> Joseph Mack mack@xxxxxxxxxxx
>
>
> ----------------------------------------------------------------------
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> To unsubscribe, e-mail: lvs-users-unsubscribe@xxxxxxxxxxxxxxxxxxxxxx
> For additional commands, e-mail: lvs-users-help@xxxxxxxxxxxxxxxxxxxxxx
>
----------------------------------------------------------------------
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
To unsubscribe, e-mail: lvs-users-unsubscribe@xxxxxxxxxxxxxxxxxxxxxx
For additional commands, e-mail: lvs-users-help@xxxxxxxxxxxxxxxxxxxxxx
|
