Scyld TaskMaster Suite has many important features and benefits. Click to expand.
- Improved Resource Utilization - Get the Most out of your Cluster Investment
Scyld TaskMaster ensures that compute jobs are started on the best suited compute nodes in a Scyld ClusterWare cluster. The placement of jobs is determined by
- Resource requirements specified by the user at job submission time
- The current availability of resources in the Scyld ClusterWare cluster
- Scheduling policies
- Resource reservations
Under consideration of these factors, Scyld TaskMaster optimizes the placement of jobs. The allocation of compute nodes for job execution is balanced across the cluster. In 'busy' clusters, Scyld TaskMaster prevents the oversubscription of compute nodes, a condition that results in serious performance degradation for individual compute jobs. Oversubscription is a common occurrence in clusters where jobs are started interactively in an 'ad-hoc' fashion, without the use of a workload manager.
With Scyld TaskMaster you get the most out of your compute cluster investment. Scyld TaskMaster's optimized job placement allows for resource utilization levels of up to 99%.
- Optimized Scheduling - Align Business Priorities with Cluster Use
While a high resource utilization level is definitely desirable, efficient cluster use is not just about resource utilization figures. In a typical HPC cluster users from multiple groups, working on a variety of projects share computational resources. The projects the users work on are typically of different relative importance to an organization's overall objectives. Scyld TaskMaster provides the industry's most flexible scheduling engine for coordinated resource sharing under consideration of an organization's priorities. A multitude of scheduling policies are available that can be combined and configured to reflect specific business requirements. While Scyld TaskMaster strictly enforces these policies, it also optimizes resource utilization and ensures that resources are not wasted. An example of a flexible policy is the preemption policy. In situations where resources reserved for high priority jobs are not used, Scyld TaskMaster's preemption policy can schedule lower priority tasks to use the reserved resources. In case the reserved resources should be needed by higher priority jobs, the lower priority jobs can be killed or suspended to 'give-way' to high priority jobs. Through a simulation mode Scyld TaskMaster makes it possible to fine-tune applied scheduling policies based on historical workload logs, without impacting a production environment.
Scyld TaskMaster offers intelligent scheduling and precise policy controls that enable you to implement scheduling policies that are in-line with your business priorities. Scyld TaskMaster ensures that jobs always run on the right compute nodes at the right time.
- Advanced Administration - Simplified Management
Scyld TaskMaster includes a task-based cluster management interface and monitoring and reporting tool. Common management tasks such as obtaining cluster status and diagnostics or staging changes can be performed faster and easier. With Scyld TaskMaster administrators can
- quickly view, modify and report on jobs, nodes, statistics and policies
- easily analyze cluster status with a visual overview of cluster workload conditions, reservations and current policies
- create and dynamically modify advanced reservations to ensure cluster availability at the right time and on the right resources
Scyld TaskMaster also supports the automation of administrative tasks through triggers. Triggers enable actions to be taken when the specified event, offset, and threshold criteria are satisfied. Triggers can be defined at a job or node level or system wide. An administrator may, for example, want to send an email to the owner of a particular reservation when the usage drops below a specific threshold.
Scyld TaskMaster's intuitive graphical management interface makes it easy for administrators to quickly evaluate the workload situation in a cluster. With the help of automated management the administrative burden of managing a scheduler is minimized.
- Integrated Reporting - Information for Cluster Tuning and Upgrades, ROI Figures or Chargeback Accounting
Scyld TaskMaster includes reporting capabilities. Custom reports can be generated to visualize information on live and historical cluster utilization, resource allocation, and backlog. Scyld TaskMaster also allows for creating reports on provided service levels and resource utilization, broken down by organizational units.
Scyld TaskMaster reports provide information about a cluster in a format that makes it easy to understand all relevant aspects. This valuable information can be used for making informed decisions about cluster upgrades, identifying resource bottlenecks and providing cluster usage information for chargeback accounting purposes or ROI calculations.
- Graphical Job Submission Interface/Job Submission Portal – Users become Productive Quickly
Scyld TaskMaster allows end-users to submit jobs to the cluster either through a Command Line Interface, a graphical job submission client or a web based job submission portal. The GUI client and job submission portal are highly intuitive, Reducing end-user job management training with an easy-to-use web interface and self-help capabilities such as start time estimates and visual reservation tools. Job submission is made easy by means of basic and advanced job forms, reusable and shared job templates and the ability to browse for needed files. Users that are used to job submission syntax of the TORQUE resource manager can submit jobs to Scyld TaskMaster using the familiar set of TORQUE commands.
With Scyld TaskMaster’s job submission interfaces cluster users become productive quickly without requiring extensive training.
Some content used by permission of Cluster Resources, Inc.