ExtremeStor B BeeGFS Appliance with Expert Design, Implementation and Support
World’s Fastest Growing Parallel File System
ExtremeStor B Delivers Extreme Performance and Scale with Easy Management
ExtremeStor B delivers maximum performance and scalability on fully integrated, top quality, industry standard hardware for maximum performance, reliability and data protection across a wide range of technical applications. ExtremeStor B with BeeGFS distributes files in parallel across multiple storage servers with dedicated, distributed metadata processing.
ExtremeStor B delivers up 8 GB/s client throughput with a single process streaming on a 100Gb network, with only a few streams capable of fully saturating the network. ExtremeStor B delivers best in class metadata performance with linear scalability through dynamic metadata name-space partitioning and distributing metadata operations per directory and subdirectory across metadata nodes in a simplified manner.
A BeeGFS deployment at UCSB delivered over 13 GB/s with RAID 6 data protection.
- 4x Object Storage Servers with 36x6TB NL SAS disks each
- 2x Metadata servers with Buddy Mirroring for high availability
- To meet performance and capacity requirements, Applied Data Systems deftly architected the appropriate drive capacity and number of Object Servers
- Optional features such as BeeOND deliver the performance of all-flash arrays as compute jobs require
- Affordable object server building blocks can be added in order to scale both performance and capacity
Engineered for High Availability, Data Protection and Fault Tolerance
ExtremeStor BeeGFS storage servers come with underlying RAID (either RAID-6 or RAID-Z2) to transparently handle disk errors. BeeGFS includes a HA mechanism that is fully integrated and which does not rely on special hardware. This approach is called Buddy Mirroring, based on the concept of pairs of servers (the so-called buddies) that internally replicate each other and that help each other in case one of them has a problem.
The built-in BeeGFS Buddy Mirroring approach can tolerate the loss of complete servers including all data on their RAID volumes – on commodity servers and shared-nothing hardware. Buddy Mirroring can also be used to put buddies in different failure domains, different racks or different server rooms.