Archived from groups: alt.internet.wireless (
More info?)
In article <cpbnj1$k8h$1@lust.ihug.co.nz>, Tim <Tim@NoSpam.com> wrote:
:You can get most of that in a D-link WAP... and you don't need the support
:'cos they just work.
D-Link might "just work" for the few situations -you- have tried, but I
have gone through hundreds of postings and web sites about people's
experiences, and the conclusion I came to was that "If D-Link works for
you, it will perform acceptably well, but if it doesn't, you will
probably *never* be able to figure out what's wrong with it and the
support will likely be somewhere between clueless and actively harmful
to your system and networks."
D-Link is, as best I can tell, a consumer class device: you
can buy it and try it out, but if it doesn't work for you in short
order, then *get rid of it* because it'll just end up wasting your time.
It's not a device to put into a business where an hour's downtime might
cost the company several thousand dollars. Cisco might be expensive,
especially their 1200 dual-radio series, but if you call in with
a "network down" problem, then within a matter of hours, you will have
someone digging through the code to solve your problem.
It's a business decision: if your network is such that you can
risk spending days or weeks off and on trying to figure out what's
wrong with your wireless connections, and can afford to try out a few
different low-end models and vendors, then and you have a lot of
patience with trouble-ticket takers who don't have much clue about
how the product works, then you can afford to work with D-Link and
other such consumer devices. But if time is money to you, and you need
the kind of support that drags a manufacturer's technician on-site
until midnight one evening, then you have to play in the big leagues.
:They have LEAP as well as WAP and the other security
:stuff (802.1x). No doubt they are more limiting in MAC's... but if thats an
:issue you can buy 4, get 4 times the throughput and 4 times the MAC
:handling.
If you are briding between two networks and the MAC table
fills up on one of the devices, it is going to fill up on the other
devices too, except now because you are working with multiple devices
you have to worry about spanning tree hastles and about the different
channels interfering with each other and the different AP's interfering
with the AP's you've installed for the floor below and the wing
across from you.
Anecdote: a few years ago, when we were making the transition to
100BaseT from 10BaseT, a 24 port 10/100 switch from Nortel cost about
$5000, and a 24 port switch from SMC cost about $1000. We did the
math, said "The SMC can't possibly be 5 times as bad", and so went
with the SMC, trying one at first, with visions of saving literally
tens of thousands of dollars when we went to upgrade all our switches.
And everything was fine with the SMC at first, until one day we
realized that some of the devices couldn't reach each other, and by
experimentation realized that all the 10 Mb devices could see each
other, and all the 100 Mb devices could see each other, but the 10 and
100 couldn't talk. "Oh," SMC said after we'd already spent several days
trying to figure out what was going on, "You have to buy a $1200
software license to enable that feature. You missed the fine print in
the advertisements." Well, $2200 was still a lot less expensive than
$5000, so we paid up, and kept going, and everything seemed to be
going fine.
But I noticed little problems here and there whose only common element
was the SMC, and although there was never anything as gauche as
complete port failures, over the next several months, I underwent "The
Death of 1000 Cuts", trying to track down why people sometimes could
not get to the mail server, and having no joy. Eventually I tired of
the unproductive late nights, convinced my boss that paying me overtime
to track the issues was not a good use of my salary or my mental
energies, and we got in the $5000 Nortel switch. And Bingo! the
problems disappeared, and the error rates I could measure fell by a
factor of 100 and I was able to do remote remote management instead of
having to drop in our network probe inline and hope for the best.
It would have been better if the SMC device had been widely faulty:
then it would have been obvious that it needed to be torn out. But
because it mostly worked, I lost a lot of nights trying to figure
out where on our network the problems were and what could be done
to fix them. What did other users have to say? "The SMC works
great for me!"
In our production network, the SMC proved to be a false economy -- and
there wasn't anything in the technical specs that would have allowed
us to know that. We had to live it and get burned for ourselves.
Now these days I'm about to have to fight that fight all over again:
unmanaged 5-port gigabit switches are ~$100 and quality 24 port
gigabit switches are ~$3000. ~$3000 is enough to pay a postdoc for
a couple of months, whereas it does't cost the researchers anything
if -I- have to suffer through long nights trying to figure out what's
going on. It doesn't cost -them- anything if I barely get to see my
spouse during the week, and it doesn't cost -them- anything if I
start burning out again. Sometimes a buck is just a buck, and
sometimes it isn't. When it comes to networking equipment in a
production environment, the -real- cost difference between
consumer equipment and professional equipment is huge, and it's the
consumer equipment that is the poorer value by far.
--
Oh, yeah, an African swallow maybe, but not a European swallow.
That's my point.