network thruput

Wed Jun 1 08:15:31 PDT 2005

>and a little script that anybody can run to show that transfering
>files on a network free of netbios is 5x faster but something
>musta been wrong with the test script..etc.etc.. it can't be .. etc..etc..
> 
>  
>
>>So, all machines have a minimum of two NICs?
>>    
>>
>
>some machines have 4 nics.. all going to the same nortel switch 
>( same 1u box ) and connected via is stackable cable in the rear
>
>some are channel bonded .. most are not
>  
>

One quick suggestion. quadruple your # of switches. I've built beowulf 
clusters where each server has
4-8 nics, each connected to a different switch, and then bonded, 
creating a 400 or 800mb (in your case
4gb or 8gb) network.  If you have the spare hardware, it might be worth 
staging that over a weekend
to see if there is an improvement. If there is, then that might not say 
that you were maxing out bandwidth,
but it might say that you were maxing out the cpus on your switches.

When you say they're getting a paltry 5-10MB/s at best, are you saying 
all of the servers are at the
same time, or at any given time?

Beyond that, the stacked switch setup could be bad if that means switch 
10 has to traverse all of the
other switches in order to get to switch 9.  If that's the case, you 
might actually want to google
for "hypertorroidal mesh" or "beowulf mesh" which will describe a couple 
of ways to build a network
with multiple nics to lower the # of switching hops from any server to a 
percentage of any other
server.

Another thing I'd do is collect some good stats to show to the PHB's .. 
Setup NTOP for a week and
show them that it's windows chatter eating up all the bandwidth. If 
they're manageable switches,
setup cacti to graph them via snmp. Might also be worth digging in to 
see if you're having any
type of arp or broadcast storms, perhaps a screwed up vlan.

> 
>  
>
>>With your above idea, why not have the backup server on both LANs? 
>>    
>>
>
>backups of 500GB during the day kills the network .. which it
>does when it goes to the $150K 10-tape library thingie 
>	which in itself should be on a separate tape lan
>
>  
>

$150k? Ouch. For $20k nowadays you can get a 40 tape lto2 library that 
has 200GB (uncompressed)
tapes.. Another $5-$7k you can go with the latest SDLT that's 300GB 
uncompressed (which I'm
looking at purchasing this week, hence the numbers in my head)

>in my world, backups should be sorta invisible on the network,
>and see if the disk latency is tolerable for head seeks
>
>  
>
I'm starting to give up on Tape to be honest. The value of tape and disk 
always goes back and fourth,
but for a new infrastructure, where you're not in-bed with a backup 
software, I'm not sure if tape
is always the right solution, excepts perhaps as an archive. I'm 
considering putting a small san
in each of my datacenters, then coming up with an alternated backup 
scheme for all of our
servers, and taking an archive of a snapshot every other day, maybe just 
once a week. Disk is just
so fast, and getting so cheap.. It's like $10k for a 6 tb iscsi raid 
from apple, that's pretty appealing,
especially with a good volume manager.

>--- 
>
>just thought it was fun ... "upper-level managers are always right"
>andits their $$$ and budget etc..
>
>c ya
>alvin
>
>  
>

-- 
-------------------
BitPusher, LLC
http://www.bitpusher.com/
1.888.9PUSHER
(415) 724.7998 - Mobile