[olug-colo] The Legal Mumbo-Jumbo

Phil Brutsche phil at brutsche.us
Mon Apr 6 17:50:27 EDT 2009


A 4912G is a 1.5U device, so basically 2U. It's not too horribly big.

Redundancy: I don't think it's feasible at this point in time.

IMO the only way redundant switches would be feasible would be to have a
pair of L3 switches at the entry point from the ISP with support from
the ISP, in the form of dual internet feeds - redundant aggregation
switches don't help any if you have a single point of failure further
upstream, and redundant routers don't help any if only one can be
connected to the ISP at once.

I would envision infrastructure redundancy like so:

Redundant L3 switches connected to the ISP, each with it's own 'net
connection.

The redundant L3 switches would then feed redundant 4912Gs. Each 48-port
switch would have a fiber link to the each 4912G.

The Catalyst 3550s support Cisco's HSRP (hot standby router protocol),
so it's at least feasible there. IEEE spanning trees make redundant
aggregation switches a snap.

Since we'll be operating - at least initially - on a shoestring budget
and on a best-effort basis I think the best we can offer is cold spares
and same-day recovery from hardware failure.

Dan Clough wrote:
> Those 4912G's look great.  As long as it's light on space it will do
> the trick.  Ten distribution lines to the ten racks is all we'll
> really need for quite a while.  When we get to the point where 24Gbps
> of switching capacity just isn't cutting it and we're overloading the
> switch, I'd think we'd have enough money on hand to beef up the
> network considerably. ;)
> 
> Would it be too complex or unfeasible at this point to consider a
> redundant network setup?  IE, main router feeding dual 4912G's, in
> turn feeding both GBIC slots on the edge switches?

-- 

Phil Brutsche
phil at brutsche.us


More information about the olug-colo mailing list