This page contains outdated information. It continues to exist as a reference document.
DataSynpase is mainly a CPU Scavenging product. Each compute client has a broker and it connects to a machine with a LiveCluster? Engine. Live Cluster run on Windows and Linux. Recommended to run the same number of engines to processors. Suggested to create groups for each type of machine. There is a single management console via http. Has shareable administration. Targets financial services, on trading application and analysis. Has lots of small jobs rather then a single large jobs. A “Run-no-Run” kind of product, no intermediate. Is based on “Pull” Technology. Some form of monitoring available.
Notes
A very Java interface and API. Has a very simple and easy to install client. DHCP usage is possible. Auto push to client is possible. Can create criterias for job submission using a driver (p-driver) using a script to specify. Use Java and C++ API to better create a HPC environment using DataSynpase.
Broker can automatically resubmit on job return failure based on timeout s set by the administrator.
Limitations
Cannot manage each box per-se, so every one gets the same data. Can only have a CPU Grid, but have a notion of file replication. Cannot have multiple version on the same grid. Cannot control who enrolls to the grid. Lack of security control, no passwords and authentication. Priority does not work. Logging is provided, but it’s kept in unstructured log files.
Queuing theory cannot be adjusted, Also, does not provide preempt/suspend for high-prioritized job which was submitted later.
Replication stages the files in a particular directory and allows the engine to copy the files from the server. Files are not encrypted. On current it is Apache and Tomcat, would be moving to Websphere and BEA weblogic.
Evaluation
Nice monitor. Permission can be by function. All GUI, no CMD line. Can input characteristics to machines and can be transferred. Almost everything can be controlled via GUI. Easy packaging for client without interaction required. Fast and easy. Need to build separate packages for each platform. Need to use their APIs.