Skip to Content.
Sympa Menu

discuss - Re: [opennic-discuss] Minimum hardware spec' for a T2?

discuss AT lists.opennicproject.org

Subject: Discuss mailing list

List archive

Re: [opennic-discuss] Minimum hardware spec' for a T2?


Chronological Thread 
  • From: "Niels Dettenbach (Syndicat IT&Internet)" <nd AT syndicat.com>
  • To: discuss AT lists.opennicproject.org
  • Subject: Re: [opennic-discuss] Minimum hardware spec' for a T2?
  • Date: Sat, 18 Feb 2012 09:21:15 +0100

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256



Peter Green <peter AT greenpete.co.uk> schrieb:

>dedicated T2
>but am concerned about the RAM not being enough, 64MB burstable to
>128MB.

I did not know V.P.S.s infrastructure, machines and hosting politics, so i
will answer in a general way here...


>I gather that BIND is relatively light weight and doesn't need much to
>run on, but
>what do you guys (and girls???) think?

This highly depends from the amount of zones that bind has (possibly) to
handle (and/or to cache) - depending from what you use the bind for at all -
and the amount of requests per seconds you want to serve with and (not at
least) if you run more then your bind beside in that VM.

128 MB could be (much) more then enough (less then 128MB i did not tried yet
in the past), but there are other often more important factors when using
somekind of VM product from any eco hoster, i.e. i/o troughputs, CPU
capacities behind, the underlying virtualization infrastructure and even
bandwidth and bandwidth quality. From my experience most VMs available from
eco hosters has very limitited capacities beside that what they write onto
their products and use relatively "poor" hardware per user/product, very
often just cheaper PC based hardware. Many hosters i've found are not deeper
experienced with their virtualization product, using suboptimal
configurations. Beside that, full virtualizations are comparably inefficient.

You could be "lucky" if he uses i.e. at least standard entry servers like HP
DL1xx or similiar IBM machines as hardware. HP DL3xxx or 5xxx (do not know
current IBMs product line) could render more throughput aso.. and even Dell
is acceptable...

Don't calculate on "bursts" as you will did not get them any / most time you
need it and the burst part could not be usable part for any FS buffers. As
i/o often is main bottleneck your binds data/buffers should "fit into" the
dedicated RAM. Avoid swapping at all...


So the only thing you can do:

- - look for a VM product which is ideally described in very detail
(hardware, manufacturer/model, maximum usage etc. and - same important -
details about the IP upstream). Loyal hosting ISPs are transparent there).

OR

- - try it!
means: install your target setup, run and/or benchmarking it and monitor it
in details like Load, Response times, Mem details, i/o statistics etc.

But be aware of the results from timely limited demos before a buy as i know
from different providers that this products could not be compared with what
you get after buy in many cases, for different reasons...

And - not at least - things may (and usually will!) change any time in the
future if i.e. new/other customers gets onto the same machine or change their
usage profile/load profile or the ISP changes underlying hardware etc...

btw: dnstop could be a very helpful tool to view into the current work of
your bind. Monitoring/ressource graphing should be done on another machine /
from outside / remotely.


good luck!
best regards,


Niels.




- --
Niels Dettenbach
Syndicat IT&Internet
http://www.syndicat.com
-----BEGIN PGP SIGNATURE-----
Version: APG v1.0.8

iIEEAREIAEEFAk8/X3o6HE5pZWxzIERldHRlbmJhY2ggKFN5bmRpY2F0IElUJklu
dGVybmV0KSA8bmRAc3luZGljYXQuY29tPgAKCRBU3ERlZRyiDVCuAJ928kV4oRES
yHuA2pTCIWqvV78XzACfZ+tVcArDa+lvtcwiM6Y7FShTkTg=
=JDCH
-----END PGP SIGNATURE-----




Archive powered by MHonArc 2.6.19.

Top of Page