WorkloadAnalyzer™ is the industry's only IT workload simulator allowing for intelligent compute farm planning. Taking a job scheduler's workload (NetworkComputer™, LSF™) along with the configuration of the compute farm (hardware & software resources) as input, a workload simulation can uncover advanced analytics of a particular compute farm configuration. More specifically, important statistics of resource utilization can be discovered, whether it be underused software licenses or overused server machines. Since WorkloadAnalyzer™ is a simulator, hypothetical scenarios can be simulated, providing CAD Managers with valuable forecasting capabilities for understanding how different resource configurations would affect the processing of a workload. Whether used daily/weekly for compute farm tuning, or quarterly/annually for software license remixes/negotiations, WorkloadAnalyzer™ is a powerful compute farm analysis tool.
The Problem
In today's times, determining the optimal resource configuration of a compute farm/cluster is a complicated problem. It depends on many variables, such as :
- Workload characteristics
- Number of HW resources available
- Number of SW resources available
- Number of company sites
- Company-specified job scheduler policies
- IT/Engineering budget
How should you mix your licenses so that the least amount of wait time is experienced by a job? Which HW resources are over-utilized or under-utilized, such that they should be either duplicated, replaced, or retired? If you are forced to cut your budget by 30%, which licenses should be the first ones cut, whereby decreasing their counts would produce the LEAST amount of newly created job wait time? Should you have $200K to invest, which HW or SW resources should be purchased first, resulting in the largest reduction to overall job wait time? If you are planning to hire a senior level HW Design Engineer and two Verification Engineers, how would their additional workloads impact the compute farm/cluster throughput?
All these are important questions that have been typically answered by human intuition in combination with homemade spreadsheets and database viewers. Unfortunately, once the size of a problem (in terms of variables) grows to a certain point of complexity, you cannot trust human intuition. You need to focus on the facts, which in this case is simulation data. Don't get fooled by thinking that increasing the availability of a license will always speed things up. Removing a computer from your compute farm/cluster will not always slow things down. Complicated problems require sophisticated solutions!
The Solution
Runtime Design Automation is excited to offer WorkloadAnalyzer™, the industry's only simulation-based workload analysis and compute farm/cluster planning tool. This revolutionary tool takes historical workloads from NetworkComputer™ or Platform Computing's LSF™, and performs in-depth analysis on the workload. More specifically, WorkloadAnalyzer™ models a particular compute farm/cluster configuration, and simulates the workload over this model. This simulation-based flow allows for the following results and capabilities :
- HW and SW Utilization Statistics
- Utilization Statistics based on Users, Groups, Hosts, Host Groups, Projects, Queues, and Licenses
- What-If Scenario Exploration
- Sensitivity Analysis
- Optimization
- Workload Editing
The true power of the tool lies in its Sensitivity Analysis Directed Optimization capability. Sensitivity Analysis Directed Optimization tells you which licenses or machines to purchase (or cut), and by how much, given a specific budget, such that doing so would produce the greatest decrease (or smallest increase) in total wait time. Scenario Exploration allows one to see what is the impact of introducing 10 more licenses of a tool, or removing a group of old machines, or increasing company-defined user limits, etc. Workload Editing changes the workload by scaling, shifting, reproducing certain jobs, based on users, projects, host groups, and so forth. This becomes a valuable feature in combination with Scenario Exploration to determine how hiring or laying off certain people, or introducing a brand-new product cycle, will affect the overall compute farm/cluster throughput.
The Conclusion
WorkloadAnalyzer™ is an extremely powerful analysis tool for any CAD Manager, VP Engineering, or CFO. Once again, complicated problems require sophisticated solutions. Take the guesswork out of compute farm/cluster planning with WorkloadAnalyzer™.
