|
On 2003-09-24T15:38:07,
Horms <horms@xxxxxxxxxxxx> said:
> LVS to be. How often does the system go down. And when it does
> what is the cause - if it is caused by something unrelated
> to LVS like faulty hardware then you can skip that bit.
In my ISP & ISP consulting times, I had been using LVS on a few (I think
it started out with four times two or so) systems as far back as 1999
(with kernel 2.2 ;), and I have never had a LVS related outage. The
systems had all been rock solid even compared to other Linux systems;
LVS load-balancers mostly exercise rather well tested code paths, while
a normal system (with disk IO load) is far more likely to run into
races.
Since working for SuSE, I never had a LVS bug assigned to me.
Most of the systems are IA32, various kernels from 2.2 to 2.4. But I
know that even on S390(x) and i/p-Series, people are using LVS as
frontends, and I've not heard of a real bug from them either.
Ranges vary widely from 4 to 32 real servers. I don't have bandwidth
data. Other than from those very initital deployment in 1999, which saw
~1k - ~5k concurrent connections, which is probably laughable compared
to todays standards ;)
So I can conclude that LVS is most certainly one of the more solid
pieces of code I have ever encountered.
Good luck with your presentation!
Sincerely,
Lars Marowsky-Brée <lmb@xxxxxxx>
--
High Availability & Clustering ever tried. ever failed. no matter.
SuSE Labs try again. fail again. fail better.
Research & Development, SUSE LINUX AG -- Samuel Beckett
|
