General |
grid*Mathematica* Local |
grid*Mathematica* Server |
Wolfram Lightweight Grid Manager

Q: Will grid*Mathematica* make all my *Mathematica* programs faster?

No, it will only speed programs written using the parallel functions in the

*Mathematica* language, and only tasks suitable for parallelization. See the

tutorial for details. Also, some built-in

*Mathematica* commands will make use of grid

*Mathematica* if it is available.

Q: What are the differences between grid*Mathematica* Local and grid*Mathematica* Server?

grid

*Mathematica* Local is always attached to an existing single-machine

*Mathematica* license and can be run only on that machine by the user of that license. grid

*Mathematica* Server kernels can be run on any available hardware, accessed from any

*Mathematica* license on the network, and be divided among any number of users up to the maximum number of kernels available.

Q: Can *Mathematica* use grid*Mathematica* Server and grid*Mathematica* Local at the same time?

Yes. A

*Mathematica* parallel program will simultaneously distribute tasks seamlessly across all available local and remote computation kernels, including any computation kernels from grid

*Mathematica* Local and grid

*Mathematica* Server.

Q: Can I do parallel computations without grid*Mathematica*?

Yes. Since

*Mathematica* 7, every

*Mathematica* license includes all of the parallel programming tools, one controlling kernel, and four computation kernels. This configuration is optimized for writing parallel computations to run on standard quad-core machines. grid

*Mathematica* scales such programs up to larger hardware.

Q: Do I need *Parallel Computing Toolkit*?

No. The parallel programming tools contained in

*Mathematica* supercede

*Parallel Computing Toolkit*, which has been discontinued.

Q: How is grid*Mathematica* priced?

grid

*Mathematica* pricing is handled on a case-by-case basis depending on your needs. Please contact

sales to receive a personalized price quote.

Q: Can I add more than four grid*Mathematica* Local computation kernels to my *Mathematica* license?

Yes, you can add as many as you want. grid

*Mathematica* Local is available in increments of four computation kernels, with no upper limit.

Q: What is the optimal number of computation kernels for my hardware?

This depends on the details of your programs, but a rough guide would be to have one computation kernel for each available CPU core.

Q: Can other people use my computation kernels from their *Mathematica* programs?

No. Your single-machine license, and any grid

*Mathematica* Local added to it, must either be used directly at the computer it is installed on or remotely by you, the registered user.

Q: Can I add grid*Mathematica* Local to a network license of *Mathematica*?

No, grid

*Mathematica* Local is only available for single-machine

*Mathematica* licenses.

Q: Do I still need *Mathematica* if I have grid*Mathematica* Server?

Yes. grid

*Mathematica* Server computation kernels are controlled by a

*Mathematica* controlling kernel. Every user of grid

*Mathematica* must have access to a

*Mathematica* license to be able to submit tasks.

Q: Can I access grid*Mathematica* Server from a single-machine license of *Mathematica*?

Yes, both single-machine and network licenses of

*Mathematica* can send computations to grid

*Mathematica* Server.

Q: Can other people use my grid*Mathematica* Server?

Yes, if you want. Each grid

*Mathematica* Server computation kernel can be acquired by a different user. This allows up to 16 people to benefit from each grid

*Mathematica* license at the same time. Wolfram Lightweight Grid Manager makes this easy and lets you control access.

Q: Do controlling and computation machines need to share a common filesystem?

No. The

*Mathematica* parallel programming tools can send any kind of data, including the program to be executed, over the connection between the controlling and computation kernels. There is no need to copy data or the distributed program onto the filesystem of the computation machine.

Q: Can I control which computers the computation kernels run on?

Yes. As long as you have the means and privileges to launch

*Mathematica* on those machines (

*e.g.* having Wolfram Lightweight Grid Manager installed), you can configure how many kernels to run on which machines. Even if you have no access to remote machines, you can acquire kernels from grid

*Mathematica* Server to run on your local machine.

Q: What grid-management systems does grid

*Mathematica* Server support?

grid

*Mathematica* Server supports Wolfram Lightweight Grid Manager, Altair PBS Professional, Microsoft Windows Compute Cluster Server, Microsoft HPC Server, Platform LSF, and Sun Grid Engine. See also the

supported platforms page.

Q: How do I set up grid*Mathematica* Server?

Explicit instructions, additional guidance, and screencasts are available

here.

Q: Can Wolfram Lightweight Grid Manager control a group of separate computers?

Yes, but Wolfram Lightweight Grid Manager must be installed on every computer in the group.

Q: Can I control who has access to each machine?

Yes, you can control which IP addresses or ranges can launch kernels via Lightweight Grid Manager. Other grid-management systems may provide other options.

Q: Can I control the machines' workload?

Yes, you can specify a maximum number of computation kernels that can run on each machine.

Q: Do I have to use Lightweight Grid Manager?

No. If you are running a

supported grid-management system, you may use that instead of Wolfram Lightweight Grid Manager. Even if you do not, you can set up your own solution for remotely launching and connecting to

*Mathematica* kernels.

Q: Are there security considerations for running Lightweight Grid Manager?

Yes. Lightweight Grid Manager allows users to run any

*Mathematica* code on your computer.

*Mathematica* can read, write, and delete files and can run other programs. You should consider who to allow to do this and what access privileges to give them. Wolfram Lightweight Grid Manager configuration control is password protected.