[kwlug-disc] Fw: Looking for colo facility suggestions

Cedric Puddy cedric at ccjclearline.com
Wed Nov 25 00:14:26 EST 2020


Super delayed response; generally agree with the thread -- I've got stuff
directly with DataCity, indirectly with Rack & Data, and directly with
NetFlash.  Looked at NPlus when they were very new, and ran away (don't
know how they've changed since).
- DataCity and Rack & Data are well connected, good core, and a decent
balance of easy-going and operating with rules; I'd talk to them both, and
make up your own mind.
- NetFlash is for a specific mindset - wants bank style handling and *zero*
wild-west.  See below for more description.

As far as heat and such goes, you *can* get a ridiculous amount of power
density into a single rack -- say, 3 fully populated blade server cabinets
with a couple 15k RPM storage shelves or some GPU shelves or something...
lets say you get it into the 7-12 kW dissipated power range.  ("Congrats,
you the man", etc.).  This isn't going to happen by accident.  Datacentres
vary, but typically, they come with a 120V, 15A gang plug (that's about 1.1
kW max) for a 42U rack -- many don't include any power *at all* as part of
your rack price, and your outlets are quoted "a la carte".  If you want
another 15A outlet, then your monthly costs just went up (say, +$200/mo).
So, if you are going to put in 10 kW worth of gear/rack, you are
*definitely* going to have a detailed conversation with the Colo Provider
about power and cooling.

One Data Centre that is in town that was not mentioned, but very excellent
in certain dimensions, is the NetFlash facility on Victoria St -- it's not
big, but if you want your stuff to be in a facility that runs like a swiss
watch, they're your guys.

   1. They don't let customers mount or cable anything -- you want stuff in
   or out, you supply the details, bring the gear over, and *they* mount and
   cable it.
   2. It's documented out the wahzo - every plug, every watt, every
   battery, every screw, every cable is to spec, labelled in a particular way,
   documented in their DCIM database,
   3. they monitor and log it beautifully.  4) Access is escorted, certain
   amount free during business hours/mo, and always paid by the hour
   afterhours.
   4. They keep an eye on things (if a light changes on your gear/it
   changes how it sounds in a not good way, you'll probably get an email).
   5. Of the local options, it comes closest to meeting Teir III specs
   ("redundant everything").

So, if you want "bank style" colocate and *zero* "Wild West", but you just
need somewhere between 1/4 and a few racks, then they are the best local
option I know of.  (If paperwork, documentation, and process give you
hives, for then stay the F*** away).  We've got a couple "high" assurance
clients who appreciate and have confidence that *nobody strange* is
anywhere near their stuff, who appreciate the pros and don't mind the cons
of this arrangement, but it certainly isn't for everyone.

As far as NPlus goes -- that building certainly was a dilapidated
warehouse/ex-factory the last time I was in, and the folks who were running
it at the time claimed it was a "Tier III" data center (cheerfully
affirming they were using the Uptime Institute definition:
https://uptimeinstitute.com/tiers ), but plainly it was a Tier 1 at the
time (The conversation is a bit like this in my memory: "Oh yes, we have
redundant power!"; "Q: what's that over there?" "A: That's the generator!
It'll be hooked up... /next week/" "Q: /the/ generator?  Shouldn't you have
two?"  "A: Oh, of course, we will!"  "Q: So, why does the ceiling look so
badly damaged?...").  I haven't been in there in at least 5 years, and they
were *just* putting initial overtures out at the time, so ... things may be
very different now, and that would be very good for everyone concerned --
if anyone is a current client with current insight, would love to hear it.

Also, if you have large wads of cash leaving inconvienient scorch marks on
your pockets, and/or you need up to 75 MW of power for your new
crypto-mining rig, there is a fancy facility right here in Cambridge (just
past the Loblaws distribution center on Maple Grove Road):
https://www.google.com/maps/place/ASCENT+TOR1/@43.4270751,-80.3750324,17z/data=!3m1!4b1!4m9!1m3!8m2!1e1!2s110468740494494847556!3m4!1s0x882b8bd3c1e14f4d:0xec6bc1d4ca61b14a!8m2!3d43.4270712!4d-80.3728437?hl=en-CA
http://www.ascent-corp.com/DataCenters/TorontoDataCenter-TOR1.aspx?gclid=CjwKCAiA-_L9BRBQEiwA-bm5fjcSzFanfmC0p2SOt8XEfZgiTXZ90tShqVtDVc1PmbtKTeGtOHrsbxoCzvUQAvD_BwE


-Cedric

On Thu, 8 Oct 2020 at 09:42, Mikalai Birukou via kwlug-disc <
kwlug-disc at kwlug.org> wrote:

> I am a total noob in racking servers. I long for explicit clarity about
> KiloWatts:
>
>  - What KW's your GPU rack is using as 90%-tile?
>
>  - How many KW's of heat vendor guarantees to take out, if all racks
> produce this amount of heat?
>
>  - Does vendor tell you this info?
>
> Implicit human assumption may be that total power available in a rack, if
> used, will be dissipated by cooling. Something tells me that even implicit
> assumption may be missing from SLA'a. (reminder to read 'em.)
>
>  - Do you have to space racks out, if they consume all of given power
> (e.g. GPU's)?
>
> It be nice to hear both consumed and energy and ability to absorb it in
> same nominal units, like KW's, as these are same energy per time, units of
> power.
>
> Super-cooled DC is super implicit. What are explicit numbers?
>
> (It happens that I write code that implicitly means the right thing, but
> it never convinces the machine. :) )
> On 2020-10-07 6:03 p.m., Charlie Drage wrote:
>
> Off-ish topic. Has anyone had an experience hosting a high temp multi GPU
> rack server at NPlus or any nearby data center? The problem I'm having is
> constant high temperatures with my cluster.
>
> So looking for a super-cooled DC.
>
> On Wed, Oct 7, 2020, 18:00 Jonathan Poole <jpoole at digitaljedi.ca> wrote:
>
>> Francisco is pretty awesome. I’d concur with that
>>
>> Sent from my iPhone
>>
>> > On Oct 7, 2020, at 6:58 PM, Digimer <lists at alteeve.ca> wrote:
>> >
>> > Excellent news, thanks.
>> >
>> >> On 2020-10-07 5:25 p.m., Charlie Drage wrote:
>> >> Just replied ^^. If Fibernetics took over, I'd suggest giving them a
>> >> try. Francisco is a networking guru.
>> >>
>> >>
>> >>
>> >> On Wed, Oct 7, 2020, 17:24 Digimer <lists at alteeve.ca
>> >> <mailto:lists at alteeve.ca>> wrote:
>> >>
>> >>    It's who we're strongly leaning towards. Can you elaborate on your
>> past
>> >>    experience?
>> >>
>> >>>    On 2020-10-07 1:29 p.m., Charlie Drage wrote:
>> >>> Super delayed reply.
>> >>>
>> >>> But whatever you do, do NOT host at Rack & Data.
>> >>>
>> >>> I had nothing but problems (I was there for 3 years).
>> >>>
>> >>> --
>> >>> Charlie Drage
>> >>> PGP - 4096R/C037D617
>> >>> GPG / Encrypted Communication: charliedrage.com/contact
>> >>    <http://charliedrage.com/contact>
>> >>>
>> >>> On Thu, May 28, 2020 at 3:06 PM Andrew Sullivan Cant
>> >>> <acant at alumni.uwaterloo.ca <mailto:acant at alumni.uwaterloo.ca>> wrote:
>> >>>>
>> >>>> Now this got be curious if anyone has a list of all the data
>> >>    centres in
>> >>>> the region.
>> >>>>
>> >>>> A quick search found this list, which has some:
>> >>>> https://www.datacentermap.com/canada/waterloo/
>> >>>>
>> >>>> Which includes
>> >>>> * DataCity
>> >>>> * N Plus Networks
>> >>>> * NeuStyle - Wat1
>> >>>> * Rack & Data
>> >>>>
>> >>>>
>> >>>> Some additional search also found:
>> >>>> * http://kwdatacentre.com/
>> >>>> * https://cloudatcost.com/
>> >>>>
>> >>>> Are there any others?
>> >>>>
>> >>>> Andrew
>> >>>>
>> >>>>
>> >>>> On 2020-05-25 7:13 p.m., Paul Nijjar via kwlug-disc wrote:
>> >>>>>
>> >>>>> For some reason Madison's post got discarded too. Sigh.
>> >>>>>
>> >>>>> Madison: Have you looked into Datacity?
>> >>>>>
>> >>>>> - Paul
>> >>>>>
>> >>>>> ----- Forwarded message from kwlug-disc-bounces at kwlug.org
>> >>    <mailto:kwlug-disc-bounces at kwlug.org> -----
>> >>>>>
>> >>>>>
>> >>>>> Date: Mon, 25 May 2020 16:59:55 -0400
>> >>>>> From: Madison Kelly <mkelly at alteeve.ca <mailto:mkelly at alteeve.ca>>
>> >>>>> To: KWLUG discussion <kwlug-disc at kwlug.org
>> >>    <mailto:kwlug-disc at kwlug.org>>
>> >>>>> Subject: Looking for colo facility suggestions
>> >>>>>
>> >>>>> Hi all,
>> >>>>>
>> >>>>> � Anyone have experience with colo facilities?
>> >>>>>
>> >>>>> � We're looking of 1/4 ~ 1/2 rack, and a facility with
>> >>    multiple internet
>> >>>>> connections and redundant power.
>> >>>>>
>> >>>>> Thanks for any advice!
>> >>>>>
>> >>>>> Madi
>> >>>>>
>> >>>>> --
>> >>>>> Madison Kelly
>> >>>>> Alteeve's Niche!
>> >>>>> Chief Technical Officer
>> >>>>> c: +1-647-471-0951
>> >>>>> c: +1-647-501-5200
>> >>>>> https://alteeve.com/
>> >>>>>
>> >>>>>
>> >>>>>
>> >>>>> ----- End forwarded message -----
>> >>>>>
>> >>>> _______________________________________________
>> >>>> kwlug-disc mailing list
>> >>>> kwlug-disc at kwlug.org <mailto:kwlug-disc at kwlug.org>
>> >>>> https://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>> >>>
>> >>> _______________________________________________
>> >>> kwlug-disc mailing list
>> >>> kwlug-disc at kwlug.org <mailto:kwlug-disc at kwlug.org>
>> >>> https://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>> >>>
>> >>
>> >>
>> >>    --
>> >>    Digimer
>> >>    Papers and Projects: https://alteeve.com/w/
>> >>    "I am, somehow, less interested in the weight and convolutions of
>> >>    Einstein’s brain than in the near certainty that people of equal
>> talent
>> >>    have lived and died in cotton fields and sweatshops." - Stephen Jay
>> >>    Gould
>> >>
>> >
>> >
>> > --
>> > Digimer
>> > Papers and Projects: https://alteeve.com/w/
>> > "I am, somehow, less interested in the weight and convolutions of
>> > Einstein’s brain than in the near certainty that people of equal talent
>> > have lived and died in cotton fields and sweatshops." - Stephen Jay
>> Gould
>> >
>> > _______________________________________________
>> > kwlug-disc mailing list
>> > kwlug-disc at kwlug.org
>> > https://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>>
>>
> _______________________________________________
> kwlug-disc mailing listkwlug-disc at kwlug.orghttps://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>
> --
> Mikalai Birukou
> CEO | 3NSoft Inc.
> _______________________________________________
> kwlug-disc mailing list
> kwlug-disc at kwlug.org
> https://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>


-- 
| CCj/ClearLine - Hosting and TCP/IP Network Services since 1997
| 118 Louisa Street, Kitchener, Ontario, N2H 5M3, 519-489-0478x102
\________________________________________________________
  Cedric Puddy, IS Director cedric at ccj.host
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://kwlug.org/pipermail/kwlug-disc_kwlug.org/attachments/20201125/a378244f/attachment.htm>


More information about the kwlug-disc mailing list