[c-nsp] HA-7600 vs. non-HA 7600

Bill Wichers billw at waveform.net
Mon May 16 20:03:38 EDT 2005


> The thinking is that the upstream circuit and provider is the weakest link
> and
> more likely to fail than any other hardware component such that it doesn't
> make sense to add redundancy to the box.
>
> Any comments or feedback?

It's a gamble. I would myself try to argue for some redundancy in the
switch, or spare parts (which may be better depending on your tolerance
for varying periods of downtime). If the upstream fails, you make an angry
phone call. If your switch fails and you can't bring it back *you* get
yelled *at*. My expierience (with customers, we do consulting :-), is that
the idea of saving money up-front always goes over well, but when
something fails later the fact that you recommended something better
doesn't matter and the customer (boss') argument is that you should have
made a better argument for better equipment when you put the system in. It
is never the customer's fault in the customer's eyes.

I would argue for a small switch chassis with a spare sup, spare gigE
card, and maybe a spare GBIC or two if you need those. I've seen GBICs
fail before, a few "regular" port blades, and very rarely a supervisor.
You do want to get at least dual power supplies though since that has
advantages beyond just protecting against the failure of a power supply in
the unit.

If you have multiple providers you can make the argument that it is
unlikely both will be down at once, and thus your equipment is the weakest
link. Ultimately you have to determine how much downtime you can afford,
and whether or not the higher cost for more/better hardware is worth the
lower probability of failure it provides.

What I usually tell our customers is that to get each additional "9" of
99.9+% availability costs significantly more than the "9" before it.
Basically 99.99% maybe costs 10 times more than 99.9% that costs maybe 5
times more than 99%, etc. Where to make the breakover between tolerating
the downtime and spending the money to get a higher probability of more
uptime is dependent upon what your particular industry and/or company can
tolerate. A carrier probably can't tolerate much less than 99.999%
availability due to SLAs, likelihood of lost/non-renewing customers, etc.
An office furniture company (for example) can probably deal with 99.9%
availability, although maybe not for all services. It's a big gamble and
takes a lot of work, crystal ball gazing, and good guesses on the part of
the net eng to determine where the best balance is.

     -Bill

*****************************
Waveform Technology
UNIX Systems Administrator




More information about the cisco-nsp mailing list