Brain Murmurs
Guerrilla computing in the mountains of Seattle
| Overview | Applications | Security | Performance | Simplicity | Cost Effective | Easy Access |
JIVA was designed with full knowledge that it would be run predominantly on borrowed computers. To this end, JIVA has been designed to be as small an imposition on the computer user as possible.
The JIVA desktop client typically works as a simple screen saver. When the computer is not being used, the screen saver will start, and the JIVA Client will begin processing. If the user presses a key or uses the mouse the Screen Saver will be terminated immediately and will return control to the user. No time is spent attempting to do a clean shutdown of the JIVA Client as it is considered to be more important to have happy users. Only a fairly small amount of work will be lost in using this approach, as the JIVA Client application will save off its state information regularly as a part of its normal processing anyway.
In short, JIVA looks and acts just like an ordinary screen saver. The user should never have the slightest concern about interrupting JIVA. No work will be lost.
JIVA was definitely made with the happiness of the IT Administrator in mind and great pains have been taken to ensure that working with JIVA is as pleasant as possible. The goal has been to make running JIVA highly desirable, particularly when compared to running a Beowulf cluster or other, less secure grid computing systems.
The ability to download and run new applications without administrative assistance has been built into JIVA, and this useful but potentially dangerous feature has been complimented with draconian client security features to ensure a pleasant and unobtrusive experience for users and IT administrators.
Administrators need only install and perform some trivial configuration of the JIVA clients, which is done with automated installers. Command line tools have been added to JIVA to make scripting large deployments on enterprise networks as painless and as simple as possible, even when deploying to mixed networks of UNIX and Windows computers.
Once the client has been installed the system is capable of running many different types of distributed applications without further maintenance. New applications are distributed from the server out to clients automatically.
None. Why not take the afternoon off?
JIVA makes software updates painless. The latest versions of the JIVA Client kernel (core library) and the current JIVA task are stored locally on the JIVA Server. The JIVA Clients provide version information about the software they currently have to the Server at the start of most transactions. If the versions are out of date the JIVA Server responds information about what libraries need to be replaced. The JIVA Client will automatically reconnect with the JIVA Server and request the appropriate updates. Once the JIVA Client has installed the new software it will resume its work.
All of this happens transparently, as soon as the updated libraries are installed on the JIVA Server. The administrator does not need to do anything with this JIVA Clients at all.
The process is illustrated in the figure below. A JIVA Server has been set up on Marvin along with a kernel update (1.1) and a new Monte Carlo simulation task. JIVA Clients have been installed on four other workstations (Ford, Arthur, Zaphod, and Trillian). The JIVA Clients initially all have version 1.0 of the JIVA Client Kernel and the Brain Murmurs Scimark2 task. The updates are copied to the server by a company analyst, who then leaves for lunch. The IT manager is away enjoying a well-deserved vacation.
The automatic update process
In the figure, Ford has already sent an Idle message to the JIVA Server and received a code update response. So Ford is now sending the Code Update request, as directed, and is downloading the new JIVA Client Kernel update.
Zaphod has already received the Kernel update and restarted, but is still using the SciMark2 task. This time Zaphod sends a new Idle message with the current task and kernel version information and receives a "task update" message back from the JIVA Server.
Trillian has already updated the Kernel and has just received a "task update" response from the JIVA Server. Trillian calls back immediately with a "task update" request and begins download the new application library.
Arthur, who is the farthest along in the update process, has the latest client and the new Monte Carlo task. Arthur sends out an Idle message along with the current task and kernel version information and gets back some data to work on. Arthur starts processing the data immediately.
