RE: M20 vs. M40

From: Tony Mumm (tonym@netins.net)
Date: Fri Jul 06 2001 - 17:39:49 EDT


While I'm sure there is no one "right" way to do it, I still think that the
dual ssb/re scenario in a M20 has value. The reasoning to me is that a
connection, from either a customer or a trunk peering with another provider,
only connects to your network once. If you lose that routing engine/ssb in
a M10 environment, that customer / bandwidth is out of service until you can
restore with a replacement. While in M20, at least you have an automatic
restoration.

But its still up to you on how you design your network. It sounds like it
has value, but with value still comes cost.

When we talk about customers/trunks only connecting on one place, that
brings up a few single points failure beyond just the common/control. We
could lose the fpc, and so up to 4 pics out of service. Or we just lose the
pic, which in like a chan DS3 scenario could mean alot of customers are
down. So I guess when we talk about redundancy, without port redundancy,
its hard to consider it a fully redundant box. And with port redundancy,
hopefully it could span FPCs, so to provide the FPC redundancy.

I guess I can see it from core router perspective, port redundancy does not
as much as people that just overbuild their core with many redundant routes.
So losing a trunk wouldn't bother them too much, routing protocols just
select the alternate route.

Thanks,

Tony

-----Original Message-----
From: Przemyslaw Karwasiecki [mailto:karwas@ifxcorp.com]
Sent: Friday, July 06, 2001 2:54 PM
To: David Mallwitz; Joe McGuckin
Cc: juniper-nsp@puck.nether.net
Subject: RE: M20 vs. M40

But cost of fully redundant M20 will approximately
equal to cost of 2 M10s and not differ very much
from cost of two M20s with single RE and SSB.
Of course depends on interfaces.

Any way -- when one of those M10 or M20 dies
you don't need to wait 5 minutes to bring second RE
online and reastabilish all BGP sessions.

So, IMHO, it is much wiser to have 2 M10/M20 than
have an illusion of N2 redundancy given by M20
with 2 SSBs and 2 REs.

Any opinions?

Przemek

-----Original Message-----
From: David Mallwitz [mailto:dave@focaleng.net]
Sent: Thursday, July 05, 2001 5:35 PM
To: Joe McGuckin
Cc: juniper-nsp@puck.nether.net
Subject: Re: M20 vs. M40

I've had 2 SSB's, 1 RE, 1 hard drive and 1 internal flash fail, also had
1 corrupted flash card. All this over the past year among a deployment
of ~40 M20s. I don't consider this bad at all, since it's been pretty
easy to diagnose and we have on site spares. Keep in mind that even with
the redundant hardware the router will need to reboot, usually requiring
2-3 minutes to come back up (note to Juniper: use a journaling file
system so we can skip the fsck on unexpected reboots). I highly
recommend the M20.

Dave

On 05 Jul 2001 14:10:08 -0700, Joe McGuckin wrote:
>
> We're looking to purchase a juniper and the final question is which
> one should we get: the difference in price isn't that much and
> the M40 will save us the hassle of having to swap chassis when we start
> using up the PIC slots.
>
> The M20 does have provision for redundant SSB and Route Engines.
> Is that really necessary? Has anyone ever has an SSB or RE fail?
>
> I notice that Juniper has moved away from the LS120 floppy to flash
> cards
> for media distribution. Has anyone has problems with the LS120 drives?
>
> Comments, advice, etc are welcome!
>
> Thanks,
>
> Joe
>



This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:40 EDT