[151244] in North American Network Operators' Group
RE: Xirrus Wireless
daemon@ATHENA.MIT.EDU (Blake Pfankuch)
Wed Mar 14 00:46:46 2012
From: Blake Pfankuch <blake@pfankuch.me>
To: NANOG Mailing List <nanog@nanog.org>
Date: Wed, 14 Mar 2012 04:45:51 +0000
In-Reply-To: <4F5FCE35.6020906@altadena.net>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
Thanks very much to all of the useful on and off list releases. =20
I would like to also thank Ron Valdez of Vall Technologies for his very pro=
mpt sales contact as well. Very unprofessional, but nice try to cover up t=
he contact with the excuse of "simple google searches while reaching out to=
local IT firms" to find my contact information and directly attempt to mar=
ket a product which I just recently asked about here, and conveniently he h=
appens to be a Xirrus Gold Partner. =20
Summary of what I have learned, including quotes from a few people who said=
it better than I can reword it. "Conceptually, it sounds like a good idea=
to increate spectral bandwidth, but I have a hunch that it falls down some=
what in practice." Several people have mentioned that only a limited numbe=
r of radios within each device (3) can do 2.4ghz at the same time (which ma=
kes sense) due to signal conflict and the specified specs which say 120 deg=
rees of broadcast per antenna. Several people have also stated (as well as=
math) that a single device can only handle about 90-120 2.4ghz clients bef=
ore there is noticeable slowdown. 5ghz wise experience holds up to specs a=
s far as client connections. Having 802.11b enabled anywhere has had a ver=
y negative on performance of the device as one could expect. In buildings =
with many smaller rooms, using a single device to cover so many rooms runs =
you into the problem of interference thanks to walls, refraction and materi=
al conflicts. Scaling them back becomes tough because each device with its=
large number of radios saturates the spectrum, allowing limited overlap..=
. "Xirrus is overkill [...] when doing small gigs and won't scale [to] ver=
y big events, compared to a truckload of cisco APs. Mostly because our venu=
es are not stadium sized." "Turn up the AP count, turn down the signal str=
ength fill the building 'til it glows."
Thanks for all the input!
-----Original Message-----
From: Pete Carah [mailto:pete@altadena.net]=20
Sent: Tuesday, March 13, 2012 4:46 PM
To: Blake Pfankuch; NANOG Mailing List
Subject: Re: Xirrus Wireless
On 03/13/2012 03:35 PM, Blake Pfankuch wrote:
> Thanks Pete, that does help. Now hopefully I can get someone who has exp=
erience with 500+ devices running on a single one in a fairly small area (H=
igh School Gym).
There was a thread about this a couple of months back, I'm pretty sure it w=
as after last November (but not absolutely sure); lots of discussion about =
density and Xirrrus was mentioned. My personal experience with Xirrus is c=
ertainly not high-density, and the "real" hospital certainly copes with a b=
unch (though I'm guessing 20-30 users per AP from how many APs they have di=
stributed among rooms. They seem to do a bunch of their device telemetry o=
n 802.11 but there are also some more dedicated frequencies/protocols for m=
edical devices. (even the IV pumps alarm at the nurse's station...)
I do have some experience with full-duplex RF transceiver design, though, a=
nd the Xirrus configuration can't be easy to make work well.=20
Not impossible, but difficult.
-- Pete
>
> -----Original Message-----
> From: Pete Carah [mailto:pete@altadena.net]
> Sent: Tuesday, March 13, 2012 4:32 PM
> To: nanog@nanog.org
> Subject: Re: Xirrus Wireless
>
> On 03/13/2012 02:34 PM, Blake Pfankuch wrote:
>> I know this is a little outside of the traditional NANOG realm but...
>>
>> I have a customer looking at a fair number of Xirrus Wireless Arrays for=
802.11a/b/g/n implementations and am looking for some real world insight i=
nto them. On the cover they look cool, the white papers look cool, but I a=
m yet to find technical commentary from a real person on these devices. Lo=
oking at the XN line, and just curious if anyone has deployed these, suppor=
ts these or knows anything about them.
> I can only speak from indirect experience; the rehab place where my=20
> wife is staying for a bit uses 4 or 5 of them (older, probably not=20
> current, flying-saucer-like boxes suspended from the ceiling at=20
> hallway
> junctions) and there, at least, they appear to work pretty well. The par=
ticular ones don't appear to my laptop to do 11a. However, I don't think t=
here is any significant user density just from watching the nifty direction=
al light display, so this may not mean much (I'd guess 3 to 10 users over =
the whole building including smartphones and a couple of pieces of medical =
equipment that isn't used much). Also there is no IT (or any real technica=
l maint) guy on-premises to talk to so I can't ask about any other aspect.
>
> The local real hospital uses a Cisco system (or at least Cisco APs; don't=
know about the AP manager box) which really does appear to work well; I'd =
guess several hundred APs with lots of full-time medical gear, and a "guest=
" network which is behind a rather draconian firewall (wouldn't let me ssh =
out to a non-standard port (65k range), for example; I had to fix myself a =
443 ssh port for the time we spent there a couple of months ago... Blocked=
25 outgoing; I don't blame them for that, however they also blocked 465 (b=
ut allowed 587)).
>
> I suspect if I wanted 2.4-only I'd go with ubiquiti, but I don't have any=
experience with them, and their "unifi" boxes don't (yet) come in 5gig. A=
nd they don't appear to have independent APs in each box, though I don't kn=
ow how well the "directional" antennas in the Xirrus actually separate thin=
gs; even a 100mw transmitter may well overwhelm all the other local receive=
rs unless there is a bunch of shielding inside the enclosure (and maybe eve=
n then...) If 802.11 was frequency-split like the cell system it would hel=
p such systems a bunch.
>
> -- Pete
>
>> Thanks!
>>
>> Blake
>>
>
>
