[kwlug-disc] VPS hosting

jpoole at digitaljedi.ca jpoole at digitaljedi.ca
Tue Nov 27 13:58:28 EST 2012


Really hard to get a 'magic' formula here as there is so many moving parts.

First off I'll weigh in on some factors..

You buy a server, be it whatever price you want to spend.

It's either going to be over provisioned (most of the time), or under
provisioned (probably not a good idea) right out of the gate.  This is the
nature of the beast.  SaaS/IaaS providers manage this, this no longer is your
concern.  This in fact is the majority of the day to day duties of a IaaS/SaaS
provider is managing capacity, moving things around, adjusting this knob, tuning
this one.

You're going to have to find a provider to host the server, all have their
pluses, or negatives. It will consume power, consume air conditioning and
consume some space, some bandwidth, require attention, etc.  Depending on what
you're running, may need drivers, patches, bios updates, firmware patches, etc
that all need to be managed by you and sometimes without the 'other' supporting
infrastructure, impossible to do/manage remote.

Depending on what your 'app' needs, you are going to be managing the scaling
process the 'old fashioned' way, of crunching numbers, gathering metrics, and
making sure you are planning for capacity.  The AWS/VPS Pay as you Go (row)
approach allows for great flexibility of 'trying' before you commit to buying,
or maybe the pay as you go is all you need.

Some frustrations I see when 'moving' to the cloud is that luxuries like
'writing logs to disk' may sound simple, just may not work the same way.  Thus,
new SaaS applications such as message queuing, or notification services (amazon
SQS, SNS).  I've had conversations with engineers that have even said to me
"Hey, what if the message queue goes down, we want to write to disk" when the
reality is, chances are if the message queue goes down, you probably won't be
able to write to disk, in which that disk may just be a 'Service' as well.  But
amazon doesn't go down, do they? (Laugh).  Morale of this tidbit is that
calculating disk IO is random at best on an ec2 instance.  Sometimes you get
lucky, but sometimes you get some random disk io activity, however I feel
amazon's disk I/O is not all that bad depending on your needs.  This link maybe
useful for figuring those out.

(http://stu.mp/2009/12/disk-io-and-throughput-benchmarks-on-amazons-ec2.html)




So I would say hosting a 'pizza' box of 1U would probably run you 80-120$/month
+ bandwidth if a service provider would even accommodate you in their
collocation environment.  I know peer1 used to have 8U racks that one could put
a server, a switch and some other components for a 400$/month price tag plus
bandwidth years ago, but not sure if they do that anymore as VPS is the better
approach for both the provider and the customer.  VPS costs really depend on
your needs, but can range from 30$/month to 3000+$/month.

> Is it worth, then, farming this out to some sort of service provider?
> Let them figure out the VPS solution - all of the issues then become
> their problem, not yours?
>
> At the very least, it would encapsulate the costs, which can be
> communicated and the powers that be then decide if they're willing to
> bear that cost. (As opposed to an inhouse server and avoiding all your
> potential VPS issues. While gaining all the issues of maintaining 'yet
> another box' - managing YAV, Yet Another Vendor, vs consumed floor space.)
>
> A service provider quote would allow you to make the point that whether
> or not they spend that amount externally, they are going to expend those
> costs one way or another.
>
> Your time has value / is a (somewhat?) quantifiable expense.
>
> Let alone the opportunity cost of what you're not doing while you're
> doing this - be it inhouse, or managing your own VPS. vs someone
> managing it for you.
>
> Anyone got service provider recommendations / ballpark costs, vs rolling
> your own remote server?
>
> Such could give Paul some ammunition.
>
> It's the same argument as to why some use java instead of writing
> everything in C, let alone assembler. At some point rolling your own is
> providing diminishing results at greater cost. Particularly opportunity
> cost.
>
> On 12-11-26 09:37 PM, Paul Nijjar wrote:
> .
> .
> .
>> It is more work and more risk, and given that we are not keeping up
>> with the work we have now, deciding to go VPS is not a trivial
>> decision.
>
> _______________________________________________
> kwlug-disc mailing list
> kwlug-disc at kwlug.org
> http://kwlug.org/mailman/listinfo/kwlug-disc_kwlug.org
>





More information about the kwlug-disc mailing list