For customers everywhere, electricity costs associated with running systems over the lifetime of the service can out-weigh the initial cost of hardware.
Concurrent Thinking announces today the extended support of power management for compute clusters and farms to expand upon the capability typically available in commodity server nodes. In particular, the concurrentCOMMAND and concurrentCONTROL appliances provide monitoring of under utilized compute resources and can power down/up servers depending on the workload volume delivered across your system. The resulting incremental savings in electricity can make a significant contribution to overall budgets as server activity can be optimized at times of reduced demand, application failure or critical service failure.
The appliances deliver the Concurrent Thinking framework that monitors the state of compute and file servers and which can react to events to maintain system health and to optimize the deployment of production workloads. One example of this is the ability to monitor the workload of individual servers against the number of jobs scheduled to run. When an imbalance occurs, events are triggered to remotely power down or power up servers.
The appliance framework supports server and storage technology from multiple vendors including Dell, HP, IBM, Sun and Supermicro, as well as network technology including Ethernet, Myrinet and QsNet. Supported software content includes both commercial and open-source tools, a wide range of middleware
options and key applications.
The availability of Concurrent Thinking's latest V3.2 Cluster Management Appliances is announced today. In addition to the above power management capabilities these bring unprecedented simplicity to the process of installing and managing HPC clusters for systems integrators and administrators and utilise new features, which include:
* Building clusters just by using the appliance
* Wizard for configuring cluster and appliance
* Simple add or replace nodes
* Stand-alone management and imaging
* Backup and restore for cluster front-ends and compute nodes
* Improved management of monitoring data
* Support for multiple Linux flavours
* Support for multiple sub-systems
These features are supported within the concurrentCOMMAND and concurrentCONTROL appliances. The appliances provide a comprehensive infrastructure for managing third-party servers through a Web 2 GUI or command line interface. Scalable imaging capabilities allow system integrators and end-users to image complete clusters in near constant time, regardless of the number of compute nodes, while an O/S independent content management system offers the ability to configure and manipulate images with pre-defined HPC content. The products also provide further system-level and component-level monitoring, with real-time and historical views of IPMI hardware metrics, usage metrics and error conditions, and scriptable responses to any breaches of user-defined conditions. Other built-in tools offer system testing and benchmarking capabilities which may be used for factory acceptance tests and problem diagnosis.