Comments inline.
JMcL
---------------------- Forwarded by Jenny Mcleod/NSO/CSDA on 23/04/2001
11:16 am ---------------------------


"ElephantChild" @groupstudy.com on 21/04/2001 04:13:51 pm

Please respond to "ElephantChild" 

Sent by:  [EMAIL PROTECTED]



To:   [EMAIL PROTECTED]
cc:


Subject:  Re: ISDN loadsharing [7:1280]


On Fri, 20 Apr 2001, [EMAIL PROTECTED] wrote:

> [ElephantChild wrote:]
>
> [You want to load-balance 30% (100% - 70% load threshold needed for ISDN
to
> kick in) of your T1's bandwidth, which is about 450 Kbits/sec., with 1
ISDN
> channel, which will give you 1/7 at best of that bandwidth.]
>
> I am not sure I understand what you are saying?
>
> Isn't the "BACKUP LOAD {enable-threshold | NEVER} {disable-threshold |
> NEVER}" command specifying when the backup interface will be kicked in,
and
> dropped?  It has been a while but looking through the docs again it seems
> that the BRI will kick in at 70% util (input or output 5 minute moving
> average).  It will remain up until the COMBINED utilization of the two
> (input or output 5 minute moving average) drops bellow 30% of the S1/0
> bandwidth.

It is. The BRI kicks in at 70% of the T1's bandwidth, so the fraction of
the T1's bandwidth that gets spread out is the remaining 30%, or perhaps
the difference between 70% and the peak usage, come to think of it. If
the original poster didn't baseline that already, now would be a good
time to.

JMcL: How much traffic will go to the T1 and how much will go to the ISDN
is a function of the routing process and protocol, not just the bandwidths.
For example, if OSPF was used, once the backup link kicked in the traffic
will be equally shared between the T1 and ISDN (per packet or per cache
entry, depending on what sort of switching), and your total throughput will
plummet.  However the original poster is planning on using EIGRP, which can
do unequal load balancing.  I have little enough knowledge of EIGRP that I
won't offer an opinion on whether this will work or not, but just on pure
bandwidth and assuming EIGRP can handle it perfectly, I don't think it will
help much - proportionally, you're not adding much bandwidth.  Depends
partly on your traffic patterns.   /JMcL

> I donn't see how the backup delay really has anything to do with the
> "loadsharing" issue?

It doesn't, but I never said it did. :-) The 30% I mentioned comes from
100% (full T1 bandwidth) - the 70 of the backup load command, not the 30
of the backup delay.

> As I write this I find that I have some questions?
> ### It has been a while, and the Docs are a little vague###

> 1. Can either the input or output utilization start the backup?
> 2. Once the backup is enabled do BOTH the input and output utilizations
have
> to be bellow the disable-threshold to drop the backup.

I don't have routers handy to experiment on, but my guesses would be:

1. No. Backup requires that the ISDN or analog line be setup to call the
   same router as the point-to-point line it's supposed to back up,
   IIRC. Hence, the output load on one router should be the same as the
   input load on the other.
JMcL: Yes.  From the command reference for the "backup load" command (my
emphasis): "When the *transmitted or received* load on the primary line is
greater than the value assigned to the enable-threshold argument, the
secondary line is enabled. /JMcL


2. Dunno either. It would make sense for it to, since otherwise the
   other router would restart the backup at once. OTOH, it assumes that
   both have the same backup load thresholds.

JMcL: I think no.  Only one router is configured to activate and deactivate
the backup.  Again from the command reference, "The secondary line is
disabled when one of the following conditions occurs:

The transmitted load on the primary line plus the transmitted load on the
secondary line is less than the value entered for the disable-load
argument.

The received load on the primary line plus the received load on the
secondary line is less than the value entered for the disable-load
argument."
Personally I think this is a bit dubious, for the reason you mention - if
transmitted and received traffic is very unbalanced, the router could
satisfy the condition to enable the backup link, and simultaneously satisfy
the condition to disable the backup link.  I suspect that it has to satisfy
BOTH conditions above to disable the link, not just one.  /JMcL


Perhaps someone would care to set up a lab and report? Chuck?
Circusnuts? Cthulu? Anyone? :-)

> "ElephantChild"  wrote:
> >
> > On Thu, 19 Apr 2001, Paul L Holloway wrote:
> >
> > > I have a customer wanting his ISDN line(BRI)to take up the slack when
the
> > > utilization on his T1 gets to high. He has a 3640 router. I know the
load
> > > balancing considerations for routing protocols and am suggesting he
run
> > > EIGRP as his IGP since OSPF won't load balance across two "unequal"
> paths.
> > > The config. on his serial I believe would be:
> > >
> > > int S1/0
> > > ip address x.x.x.x x.x.x.x
> > > no ip redirects
> > > no ip directed-broadcast
> > > no ip proxy-arp
> > > encapsulation ppp
> > > no ip route-cache distributed
> > > no fair-queue
> > > no cdp enable
> > > backup delay 30 60
> > > backup int BRI1/0
> > > backup load 70 40
> > >
> > > Are there any other snags or problems I may run into with this setup?
> > > Thanks in advance.
> >
> > Hmm, you configured it as a serial interface, wo I assume that the
whole
> > bandwidth of your T1 goes to a single destination (eg, ISP router
port).
> > You want to load-balance 30% (100% - 70% load threshold needed for ISDN
> > to kick in) of your T1's bandwidth, which is about 450 Kbits/sec., with
> > 1 ISDN channel, which will give you 1/7 at best of that bandwidth.
> >
> > So the main snag, I would think, is that your ISDN won't help that
much.
> > OTOH, it will come handy when your T1 gets acquainted with a backhoe,
> > which also appears to be one of your design goals, since you put in a
> > backup delay. Depending on how you rank these design goals, you'll get
a
> > satisfied customer or an irate one.

--
"Someone approached me and asked me to teach a javascript course. I was
about to decline, saying that my complete ignorance of the subject made
me unsuitable, then I thought again, that maybe it doesn't, as driving
people away from it is a desirable outcome." --Me
FAQ, list archives, and subscription info:
http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]




Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=1555&t=1280
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]

Reply via email to