Panasas Predictive Self-Management encompasses a number of key technologies that work together to deliver elements of health monitoring and self-healing, ensuring the Panasas Storage Cluster continues to operate and provide maximum performance to the users and applications running on the cluster. These technologies can be grouped into four major areas: monitor and fix; monitor and alert; load balancing; and redundancy.

ActiveScan detects and corrects potential media defects and preemptively moves data from a potentially malfunctioning StorageBlade module to other StorageBlade modules to eliminate the need for data reconstruction. By using statistical analysis of the disk drive SMART attributes, potential future disk drive failure can be predicted and action can be taken to protect data before failure occurs. If a disk drive failure is predicted, then the data objects can be pre-emptively migrated to other StorageBlade modules, eliminating the need for reconstruction. A replacement StorageBlade module can be inserted and the DirectorBlade modules will migrate data objects onto the new blade as part of their normal capacity balancing routine. If a StorageBlade module should happen to fail while its contents are being migrated, then only the remaining portion of the blade needs to be rebuilt. This significantly reduces the reconstruction workload and maintains performance. In addition to the disk drive attributes, ActiveScan continuously monitors data objects, the RAID parity and the disk media. If a potential problem is detected with the sectors on the disk, then the data can be moved to spare sectors on the same disk.

ActiveGuard provides high availability with comprehensive failover capabilities for DirectorBlade modules and for the storage network.

Fans, power supplies and backup battery are constantly monitored for optimal performance, alerts are sent in the case of abnormalities.

Temperature is closely monitored for maximum reliability, if the temperature exceeds safe limits then the system will gracefully shut down.

Active and passive object balancing across StorageBlade modules ensures maximum performance and avoids hot spots.

A hot client monitor watches client loads, allowing the system administrator to redistribute application loads across more client nodes.

Read coalescing automatically maximizes cache utilization for optimum performance.

All of these technologies are interlinked and support the ActiveScale operating environment in providing the highest continuous performance with a level of manageability previously unseen in clustered storage. Panasas has extended it's product offerings to support interactive applications, providing a unified HPC storage fabric across the entire technical computing workflow. This makes technologies like Predictive Self-Management even more important. Unified HPC storage systems can easily reach Petabyte scale - making reliability critical. Unlike typical SAN-based storage systems used in HPC environments, which have three separately managed components, the Panasas ActiveStor systems are fully integrated making Predictive Self-Management possible while dramatically lowering the total cost of ownership.