There is a GigE on each of our two media servers. One does 50k packets/sec
during the nighttime hours, the other does considerably less. Though it's
combined throughput is more like 25k packets/sec. That's a lot to push through
our routing core, even in off hours, if there is a better way. We could easily
have less FE interfaces and still absorb the throughput -- pipe to the server
is not the issue.
The key is to keep traffic on vlans, and therefore away from routing hops. Our
layer-2 mesh is underutilized, on purpose, for scalability. Some vlans are
combined to an interface, if there is a maximum of one router hop, and various
networking equipment parameters are observed, but only on 'low-traffic' VLANs.
With that said the true 'why do we do this?' is because the previous product we
used was designed to do this gracefully, and we could. Saves having to install
a backups VLAN, which we have to do now, at a non-trivial cost. My yearly
budget did not take into account putting a Cisco 5500 in each data center on
our campus. ;-) We have investigated VLAN tagging, which does not scale at this
size. Putting my amount of traffic through the routing mesh is also not an
acceptable solution in our environment. (as much political as technical, be
sure)
Ergo, I'm left with an inelegant solution which costs me money and other
products do. Therefore, a feature request. :)
JF
On Mon, Oct 16, 2000 at 01:39:51PM -0700, Bob Bakh filled up my inbox with:
>
> Why make it so difficult, why not replace the quad card with a Gigabit card
> and connect it to the switch, why do you want to have 4 or 5 small pipes and
> all the routing issues associated with it, and not just get a Gige port?
>
> Also if you want why not stick an f5 in front of the server and use it to
> route to the least busy port. Lastly why not just trunk the interfaces.
>
> If you have 30 v-lans then there is going to be the problem of routing them,
> so you will need 30 interfaces or a common route to all the vlans.
>
> Bob
>
> -----Original Message-----
> From: Joshua Fielden [mailto:jfielden AT excitecorp DOT com]
> Sent: Monday, October 16, 2000 11:13 AM
> To: Ravi Channavajhala
> Cc: veritas-bu AT mailman.eng.auburn DOT edu
> Subject: Re: [Veritas-bu] Network load balancing
>
>
> I just had to put in an RFE, because the software will not do this.
>
> JF
>
> On Mon, Oct 16, 2000 at 02:02:11PM -0400, Ravi Channavajhala filled up my
> inbox with:
> >
> > Hi All,
> >
> > Is it possible to do network load balancing across several
> > physical interfaces in a media server i.e., one storage
> > unit being shared by several netbackup slave server
> > definitions eg., backup-e0, backup-e1, backup-e2? The
> > primary definition is backup (nodename). All these interfaces
> > are on different subnets. The storage unit is defined for
> > backup instance only.
> >
> > I already defined these in bp.conf on the master as well as
> > the slave. What I dont understand is, how does the master
> > make a determination to direct a specific backup stream to
> > a specific network interface. Thanks.
> >
> > -ravi
> >
> > _______________________________________________
> > Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> > http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
> --
> "Any man page that includes the words "USE AT YOUR OWN RISK. BEWARE OF DOG.
> SLIPPERY WHEN WET" means trouble" - Michael Lucas
> Joshua Fielden, Senior Systems Administrator and Backups Team Lead
> eXcite@Home, Inc. jfielden AT excitecorp DOT com 650-556-3316
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
--
"Any man page that includes the words "USE AT YOUR OWN RISK. BEWARE OF DOG.
SLIPPERY WHEN WET" means trouble" - Michael Lucas
Joshua Fielden, Senior Systems Administrator and Backups Team Lead
eXcite@Home, Inc. jfielden AT excitecorp DOT com 650-556-3316
|
