When calculating more than one scenario, WEAP can calculate multiple scenarios simultaneously, each in their own separate "instance" of WEAP. Because most modern computers have multiple CPU "cores," each instance of WEAP can run on a separate core, thus achieving a significant speedup over running all scenarios sequentially on a single core. Go to Help, About to see how many CPU Cores your computer has. In addition, you can utilize other computers on a local area network as additional workers, or rent workers by the minute in the Microsoft Azure cloud. See Which to choose: your own computer, your office network, or Azure?
Turn on the Use Parallel Calculation checkbox to activate parallel calculations. (This option is not available in the following circumstances: only one scenario needs to be calculated; your WEAP area is linked to LEAP or QUAL2K; you are doing PEST Calibration or a Safe Yield Analysis; you are running the unlicensed evaluation version of WEAP.) Choose the Worker Location -- Local Only, Network or Azure. Choose the Number of Workers to use (Local, Network or Azure), up to the number of scenarios to calculate. For Network or Azure, you can optionally also start a few Local Workers. Each worker will be a separate instance of WEAP. To achieve the maximum speedup, you should set the total number of workers equal to the number of scenarios to calculate.
For the Local Only option, when calculations begin WEAP will launch several new instances of WEAP. The original instance of WEAP, the "Manager," will be moved to the left side of the screen, and each new instance of WEAP, the "Workers," will be tiled on the right side of the screen. You will be able to see the progress of calculations in each worker, as well as in the manager. Once all scenarios have been calculated, the Workers will close and the Manager will open the Results View to display the results. If you want to stop the calculations before they are complete, click the "Cancel" button. If WEAP does not hear from a new worker within two minutes, it considers that worker dormant and tries to start another one.
Note: If you have made any changes to the area, WEAP will automatically save them so that the changes can be copied to the Workers.
In cases where you have more scenarios to calculate than cores on your computer, you can recruit other computers on your local area network (LAN) as additional "freelance" workers to achieve an even greater speedup. The manager and all of the worker computers communicate through a single shared folder.
Choose a shared folder that the manager and every worker computer has read and write privileges -- a UNC path such as \\server\share\WEAP-Jobs, or a mapped network drive such as Q:\WEAP-Jobs. This folder is only used to pass messages and files between the computers; the workers do not run from it.
On the Manager, open the Parallel Calculation dialog, set Worker Location to Network, and enter the shared folder. Set the Number of Local Workers to Start to 0 to let the other computers do all the work, or to a higher number to also use the manager's own CPU cores.
On each of the other computers, install the same version of WEAP, then start it with the /Worker command-line switch and tell it which shared folder to watch with the /JobChannel switch, for example:
"C:\Program Files (x86)\WEAP\weap.exe" /Worker /NetworkJobChannel:"\\server\share\WEAP-Jobs"
(With no path -- just /NetworkJobChannel -- the worker uses the network folder previously entered or used on that computer.)
Leave that WEAP instance running; it sits idle until the manager sends it scenarios to calculate. After it finishes calculating its assigned scenarios it stays open, waiting for new jobs.
Choosing among the available workers: every worker watching the shared folder announces itself, and the manager's dialog lists them ("2 Network Workers Available" in the picture above), each with its computer and worker name, CPU speed, memory, free disk space. Use the checkboxes to choose which workers will take part in the calculation -- for example, uncheck a colleague's computer that you know is about to be shut down. Note that for the Network location this list is the only source of workers: WEAP cannot start workers on other computers itself, so someone must have started a worker on each computer as described above.
The list can briefly include workers that are no longer really available -- for example, a worker computer that was switched off without closing WEAP. The manager checks each listed worker every few seconds; one that stops answering is greyed out and marked "NOT RESPONDING", and within about half a minute it is removed from the list automatically. (If it was only a momentary network hiccup, the worker reappears in the list by itself.) Scenarios are only ever sent to checked, responding workers.
To start several workers at once on one computer, use /Worker:N to start N workers, /Worker:Max for one per CPU core, or /Worker:Ask to be prompted for how many. You can create shortcuts on the Windows desktop with these parameters to simplify starting in the future. The worker windows tile themselves automatically. When you close one worker window, WEAP offers to close all of the workers on that same computer. All participating computers must have the same version of WEAP installed.
"C:\Program Files (x86)\WEAP\weap.exe" /Worker:8 /NetworkJobChannel:"Q:\WEAP-Jobs"
With Worker Location set to Azure, WEAP rents virtual machines in the Microsoft Azure cloud, one worker per machine: it starts them, sends them your area, collects the results, and deletes them when the calculation finishes. You pay Microsoft only for the minutes the machines run. Before the first use, a one-time setup (about 30-45 minutes) creates your Azure account and everything WEAP needs in it -- see Setting Up Microsoft Azure for Parallel Calculations for detailed instructions.
Choose the Number of Azure Workers to Start and the Azure Instance Type -- the dropdown is a table comparing each machine type's memory, fast local disk, cost per hour and how many more workers your Azure quota has room for. The default type is generally the fastest and cheapest. If your Azure quota does not have room for the workers you asked to start, a message appears with a Request Quota Increase button that files the request for you -- modest increases are often approved automatically within minutes, and when Azure declines one instead, WEAP's message explains the short free support request in the Azure portal. (Clicking the button again first reports whether an earlier request is still being processed or was declined.)
Use Spot pricing (on by default) rents the same machines at Azure's deeply discounted "Spot" rate -- typically 60-90% cheaper (the type table shows both prices side by side). The trade-off: Azure may interrupt a Spot worker at any time if it needs the capacity back. This is safe -- an interrupted worker's scenarios are automatically reassigned to the remaining workers -- but on a rare busy day it can make a run slower, so uncheck the box for time-critical runs. Note that Spot workers count against their own, separate Azure quota ("Total Regional Spot vCPUs"), and increases to that quota usually require a short free support request in the Azure portal rather than being approved automatically -- the quota message will point the way when needed.
Already-running Azure workers (started earlier by you or by a colleague in the same storage account) appear in the same available-workers list as in Network mode, with the same checkboxes to choose which ones take part, and can be used in addition to, or instead of, newly started machines. This is not the usual way to use Azure -- normally the list is empty and WEAP starts fresh machines for each run -- but it is useful for repeated test runs, where keeping a machine running avoids the 3-minute startup each time. The same liveness checking applies: a listed machine that has since been stopped or deleted is greyed out as "NOT RESPONDING" and removed from the list automatically within about half a minute. When the run completes, WEAP deletes the machines it started; machines someone started by hand are left running for their owner to stop.
When your own computer (Local Only) is the right answer (try it first!): the whole calculation already finishes in under roughly half an hour, and you have as many spare cores as scenarios (many desktops have 8-16). One caveat: every local worker loads its own copy of your area into memory and competes for the same disk, so a very large model may only support 2-3 local workers before your PC runs out of memory -- and your computer will be sluggish while they run.
When Network is better: several office PCs with WEAP sit idle (evenings, a lab room) -- real hardware you already paid for, and models never leave your building. Caveats: every machine needs the same WEAP version, the shared folder can be a bottleneck for big areas, and you are borrowing computers someone may reboot.
When Azure is better: you have more scenarios than local or office cores -- Azure runs 10 or 20 workers as easily as 2, so a weekend-long run can finish over lunch. Your model is too big for several copies to fit in your PC's memory: each cloud worker gets a whole machine to itself. You want your own computer to stay fully usable while the calculation runs. Typical price: well under a dollar per large run.
When Azure is worse: short calculations -- workers take about 3 minutes to boot and must download your area first, so a 10-minute local run gains nothing by moving to the cloud. Very large areas on a slow internet connection, where the upload dominates. Organizations that do not allow cloud services or have no way to pay for one.