Mon, May 20, 2013 at 08:32:08PM CEST, fbl(a)redhat.com wrote:
On Mon, May 20, 2013 at 10:31:33AM -0700, Linda A. Walsh wrote:
> I was wondering if anyone had looked at how, say, 2 (or more)
> aggregated connections using 10Gb ethernet connections
> compared as teamed vs. using them as bonded?
>
> I know that a major hit in performance on both linux and windows
> platforms is the time spent in interrupt processing (mostly in
> deferred or SW interrupts, not HW interrupts).
>
> With teamed connections on linux being in user space, I would
> tend to think about how user-space file-system drivers perform
> compared to drivers in the kernel. On another list, people were
[...]
This is something we need to clear up in the wiki page.
The teamd is an userspace component but it is not part of
either TX or RX paths. Therefore all the heavy traffic happens
inside of the kernel. Because of that you should see little
performance impact when comparing to using without Team
or bonding driver.
As Flavio said, the hot path is in kernel and incoming and outgoing
packets never leave kernel-space. Comparing to bonding and plain 10G
dev, the overhead is pretty much the same. No penalty in compare to
bonding at all (overhead is even slightly lower)
Jiri
--
fbl
_______________________________________________
libteam mailing list
libteam(a)lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/libteam