This is a great question, and I'll be very interested to hear feedback from
others. I haven't done any load testing on the Cane River system, because
it's an in-house server and traffic has not ramped up on it yet (also it's
pretty heavy duty with tons of RAM, so not really a good test case).
For the current project I'm working on, we have budgeted for an m3.large
instance, which, if you buy at the 3-year upfront reserve price, is $1373,
or about $38/month. That's 2 cpus with 7.5 gb of memory, and with Amazon
you can set whatever size you want on the storage volume.
I have done all development on a t2.micro instances, which only have 1 gb
of memory, and have not had a problem at all. One big caveat is that when
you load a .arches file, the entire set of resources load into memory
before being saved into the database. So, I maxed out the 1gb of memory
during that process, but that was a one time operation. (with AWs you can
temporarily ramp up to a much more powerful instance and you are only
charged for the time you use, so that was an easy way to deal with that
problem at the time).
So personally I would guess that you could get away with under 8gb memory
for production, but haven't done that yet. The ability to switch sizes at
any time is one thing that makes AWS especially useful for these
applications.
Adam
On Friday, October 30, 2015 at 7:00:28 AM UTC-5, Angela Labrador wrote:
>
> I understand the minimum requirements as posted for Arches is 8-16GB of
> RAM, which for institutions with in house servers or dedicated server
> hosting levels (already), is fairly economical.
>
> But for small organizations with little to no budget, a vps or cloud setup
> would be nice. However, having 8-16GB of RAM in reserve typically scales a
> vps plan into uneconomical and fairly over spec'ed territory. For instance,
> on linode, 8GB of RAM also gets you 6 CPU cores, 192GB SSD, and 8TB
> transfer at $80/month, which seems ridiculously overpowered. With AWS EC2,
> you're looking at a t2.large instance at the least, which puts you at
> $76/mo.
>
> For those of you running Linux vps or cloud instances out there, I'm
> curious what your average load has been and what resources your instance
> needs in reserve. For those on AWS, are you on a t2 instance, and if so,
> how are the CPU credits working out?
>
> I understand this all scales depending upon dataset size and number of
> requests, so any stats people can share would be really helpful for
> estimating budgets and planning for future steps up in instance specs as
> the dataset grows.
>
> Many thanks,
> Angela
>
> P.S. I can share the following for my testing instance:
> Linode 1GB RAM 1CPU core running Ubuntu trusty with the sample Arches 3
> hip data, with 1 demo user accessing, the load was minimal: max CPU usage
> 8%, i/o rate 76 avg with highest spike at 1890, using 1% of monthly
> bandwidth to install and access (2 work weeks of fairly regular
> access/usage).
> But getting some production data would be nice...
>
--
-- To post, send email to archesproject@googlegroups.com. To unsubscribe, send
email to archesproject+unsubscr...@googlegroups.com. For more information,
visit https://groups.google.com/d/forum/archesproject?hl=en
---
You received this message because you are subscribed to the Google Groups
"Arches Project" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to archesproject+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
