NetworkComputer™ is an enterprise-grade high-performance job scheduler, otherwise known as a batch processing system or a distributed resource management system (DRMS). NetworkComputer™ provides a cost-effective solution for distributing an IT workload over any compute farm (cluster) topology.
Features
- Built-in integration with FLEXlm™ and other license managers.
- Allocation of licenses across multiple sites with MultiQueue.
- Comprehensive description of resource requirements for each job, including detailed license and hardware requirements.
- Monitors out-of-queue license usage by users, and adjusts in-queue license availability accordingly.
- Event-driven providing micro-second latency for scheduling jobs; very efficient scheduler for short jobs.
- Versatile environment management mechanism for setting up projects and tools for workflows.
- Integrated web server for user interface and your custom pages.
- Visualization of compute farm and job status and utilization.
- Comprehensive reporting of workload, waiting times, unmet demand.
- Web browser UI, CLI, and GUI provide flexible ways of using NC
- Full-access Tcl-based extension language provides excellent scriptability and automation.
Comparison with LSF™ and Grid Engine™ (previously known as SGE)
LSF™ by Platform Computing and Grid Engine™ (previously known as SGE™) by Oracle are common schedulers found in many industries. The advantages of NetworkComputer ™ over these schedulers are:
- Easier integration with FLEXlm™
- Easier installation and configuration
- Higher throughput with faster scheduler
- More complete fairshare and preemption
- Dynamic visualization of workload
- Allocation of licenses across multiple sites via MultiQueue
Scalability of the NetworkComputer Scheduler
The NetworkComputer scheduler manages farms with many thousands of processors,
with instantaneous workloads in excess of 1 million jobs, and throughput of millions of jobs per day.
At such high loads, the response time of the scheduler remains below one second and the
memory footprint of the central server remains below 2GB.
Get the most out of your resources
A chip design company had a farm with about 100 cpus and 100 simulator licenses and could not get the overall utilization of their license above about 80% using their conventional scheduler. After moving to NetworkComputer™ , they immediately reached close to 100% utilization of those licenses, thus justifying further investments in both hardware and software. They now run in excess of 1,000 licenses on a much larger farm, with sustained utilization of those licenses above 98.6%.
