|
|
| CARVIEW |
Q&A
How do I enable Wolfram Compute Services in my Wolfram product?
If you are using Mathematica, Wolfram|One or Wolfram Engine 14.3 on your desktop, run RemoteBatchSubmissionEnvironment["WolframBatch"] once to enable. The functionality is automatically available in the Wolfram Cloud now, and will be in future versions of Wolfram desktop clients.
What types of jobs can I launch?
Wolfram Compute Services supports single jobs and array jobs. A single job launches a single machine instance. An array job launches multiple machine instances determined automatically based on the size of the input array. You can specify the method by which the input array is split over machine instances. For more information, see RemoteBatchSubmit and RemoteBatchMapSubmit.
What versions of Wolfram Language are supported?
Wolfram Compute Services batch jobs can be submitted from Wolfram Language Version 14.3 or higher. The submitted jobs are always executed with the latest released version of the Wolfram Engine.
What are my options to store my computation results?
Results are stored temporarily by Wolfram Compute Services for 14 days, during which you can download and store them locally. You can also use functions like CloudPut or the AWS service connection to upload your results to the Wolfram Cloud or other cloud storage.
Can my jobs access the internet?
Yes, any Wolfram Language function that directly or indirectly accesses the internet will work as you expect. Also, all Wolfram Cloud and Wolfram|Alpha services are accessible from your jobs.
Can I SSH into my own job?
No, the batch nature of Wolfram Compute Services does not allow for SSH access to individual jobs.
How do machine classes relate to array jobs?
If you select a machine class for an array job, then that class applies to each child job. As a result, the total rate of credits consumed scales with the number of child jobs. For example, if you launch an array job that launches 16 child jobs, then each child job uses the specified machine class. If your input array has 10,000 elements, then typically 100 elements will be assigned to 100 machine instances. The machine class that you specify determines the class of each of those machine instances.
Can I use my own cloud compute account like AWS or Azure?
Yes, but not via Wolfram Compute Services. Instead, you would configure your own "AWSBatch" or "AzureBatch" provider.
What Wolfram Language functionality is not available for computation jobs?
CUDALink and other GPU-based functionality is not currently available.
Can I install custom software/tools for my job?
As part of your own job code, you can install custom software and tools as needed for your computation. Note that all customizations will be lost when the job ends.
Do I have to buy Service Credits to use Wolfram Compute Services?
Not necessarily. Many Wolfram subscription plans include a small number of Service Credits. These are enough to try small remote jobs. For larger or more frequent computations, you can purchase additional credits at any time. You can also enable auto-recharge with an amount you choose—helpful if you want your balance to stay topped up.
What happens to my jobs if I run out of Service Credits?
If you run out of credits, all running jobs will be terminated, and new jobs can't be submitted until your balance is restored. Once you add more Service Credits, you can submit new jobs or resubmit any that were interrupted.
How do I check my Service Credits balance?
You can go to your Service Credit Details page in Wolfram Account to view your balance, purchase more credits and review past usage history. You can also query your current Service Credits balance with $ServiceCreditsAvailable from Wolfram Language.
How long are my Service Credits good for?
Service Credits never expire. They remain in your account until you use them.
Can I get a refund on unused Service Credits?
No. All Service Credit purchases are final.
Does Wolfram guarantee the availability of third-party services used through Wolfram Compute Services?
No. Wolfram Compute Services uses third-party compute resources to run remote jobs. While Wolfram integrates them into the Wolfram Language workflow, we do not control their availability, performance or uptime.