Yesterday I created an EC2 account, installed Energy+, and ran a couple
of the packaged examples on a cloud machine. It took about an hour all
told to get started, mostly reading docs. I'd be happy to help if you
go this route.
This seems good for small numbers of sims. For big search problems,
next steps might be:
* package a disk image that has suitable software installed
* set up a work item queue, so that each computer automatically pulls
the next sim from the queue
* put postprocessing on the cloud machine to reduce bandwidth costs
Anyone interested?
Daniel> Keeping in mind that this is somewhat hypothetical at the moment.
On Thu, 22 Dec 2011 09:17:28 -0500, "Jim Dirkes" <jim@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> Dear Marcus,
>> . A multi-use facility (i.e., many different occupancy types and
>
> It's my understanding that E+ spends the most time figuring out the energy
> balance of all the mass in the building. Additional time, although of a
> smaller magnitude, is spent on HVAC energy balances and report variables.
> That said, I am imagining:
>
> HVAC systems)
>
> . In a roughly 300,000 m2 (3,200,000 ft2) facility
>
> . Which wants to attain LEED certification> Marcus Jones, M.Sc., LEEDRAP BD+C Non-text part: text/html
>
>
>
> I anticipate no difficulty creating HVAC systems since my past experience
> covers most or all of the systems, but the sheer size suggests that there
> will be a LOT of run time, even on a fast PC. My strategy will be to
> prepare small portions of the building and its HVAC individually, test them
> and make sure unmet load hours are acceptable. After the smaller portions
> are complete, I'll join them together for a complete building simulation and
> the various rotations.
>
> In this scenario, I should be able to keep several building portions
> underway simultaneously if I am using multiple PCs or the cloud equivalent.
> Up until this point I have been focusing on making the energy modeling
> process more effective, accurate and fast. Because I don't run full year
> simulations until the end of the process, I've been tolerating long run
> times. Now I want to take (or at least evaluate) the next step in making
> the whole process go faster.
>
> I just finished a small multi-use facility (~10,000 m2) which required 2
> hours for a complete run on a PC with reasonable speed (i5 CPU, 6GB RAM and
> solid state hard drive) - too long!
>
> At this point I think that not too many of the people who participate on
> this forum are worried about this, but I know some of you have addressed the
> issue and appreciate the feedback I've received.
>
>
>
> The Building Performance Team
> James V. Dirkes II, P.E., BEMP , LEED AP
> 1631 Acacia Drive NW
> Grand Rapids, MI 49504
> 616 450 8653
>
>
>
> From: EnergyPlus_Support@xxxxxxxxxxxxxxx
> [mailto:EnergyPlus_Support@xxxxxxxxxxxxxxx] On Behalf Of Marcus
> Sent: Thursday, December 22, 2011 4:15 AM
> To: EnergyPlus_Support@xxxxxxxxxxxxxxx
> Subject: RE: [EnergyPlus_Support] Running E+ on the "cloud"
>
>
>
>
>
> Dear Jim,
>
> What type of application are you interested in when you say "large project"?
>
>
> If you are describing a single large ASHRAE 90.1 model that may take over 15
> minutes to run, I often simply use the "group of input files" tab to execute
> my approx. 8 models (i.e. 4 baseline and 4 proposed with different
> oversizing in the HVAC) using my 4 processor cores simultaneously, so it
> takes 30 minutes instead of 2 hours. And if you do this often, it might make
> sense to buy a new dedicated multicore core machine that you can use to
> execute jobs, and access the files over a network. I would recommend this
> approach just because it requires no extra software or setup.
>
> If however you are doing global optimization or parametric studies with
> HUNDREDS of executions, then you will likely need the approach that Dr.
> Crawley or Dr. Li have mentioned, and invest into grid/cluster/cloud
> computing. Fortunately, this is a problem that has been solved many times
> over in the IT world and there are many approaches and platforms available,
> the problem really is choosing one.
>
> (Aside: This is all a topic near and dear to my PhD and professional
> objectives, so if anyone wants to submit a paper or research project let me
> know!)
>
> Cheers,
>
> Marcus
> --