Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Patents

Journal trajano's Journal: Cell computers

As you know there are a lot of portable storage devices some the size of cigars with a USB port. What if instead of a USB port, it has a 802.11g wireless stack? Then you can have something called bucket 'o storage where if you need more storage just chuck another one into the bucket.

But let's extend that concept a bit, instead of just storage, what if we used it for compute power and RAM?

For lack of a better term, I would call each compute and storage unit as a "cell"

In theory, we can have a computer in a bucket which can be infinitely scalable just throw more "cells" into it.

We can also have more specialized cells, perhaps a cell that specializes in MP3 decoding, DivX encoding, 3D raytracing, etc. Cells can also have special purpose storage for database like structures and search engines. The system would be able to delegate some tasks to these specialized cells as needed, to free up room for the general purpose cells.

We would still need a way of ensuring that each cell is allowed to be used in the bucket, so we would need some device used to do initialize each cell. I would call this the "initializer".

The "initializer" would have a socket where each cell can be put into and programmed so that it will only respond to the the network that it expects.

We also need a way for each of these cells to communicate with each other, though peer-to-peer communications is good, a centralized hub is more efficient I think, (I mean we usually use Wireless Access Points instead of ad hoc networks). For lack of a better term for this, I would call that a "spine" like the spinal cord of the human body.

Although wireless may be good, sometimes a wired connection is more cost efficient. The "spine" can also be a USB 2.0 hub whereby you can attach "cells" into it directly so it will be powered by AC instead of batteries.

How do we get all of them to work properly? Well you would still need a standard computer to talk with them because it is so much more easier. The computer does not need to be all powerful though, as it just offloads its tasks to the "bucket". To the computer, it sees the general purpose storage "cells" as one or more drives, depending on the "initializer". It sees the compute "cells" as web services where they send data to. It does not see the RAM "cells" though because those are used internally by the "bucket". It will also have a JMX/SNMP type connection to the "spine" so it can track problems such as low battery power on some of the cells so they can replace it.

The way things would work is:

  1. Controller: Ask "spine" get a free delegate cell. A delegate cell can be just another general purpose compute cell that is designated as a delegate by the spine, or it can be a special purpose cell that is programmed to perform delegate operations quickly.
  2. Controller: Gives delegate code and data to process
  3. Delegate: Ask "spine" to find appropriate compute cell
  4. Delegate: Give data and code to the compute cell
  5. Delegate: Tells controller which cell will give the information
  6. Delegate: waits for next request
  7. Compute cell: performs computation, may request other cells to help out if the request is threaded.
  8. Compute cell: sends data to the controller.
  9. Compute cell: waits for the next request.
This discussion has been archived. No new comments can be posted.

Cell computers

Comments Filter:

The only possible interpretation of any research whatever in the `social sciences' is: some do, some don't. -- Ernest Rutherford

Working...