Unison Appliances are the perfect building block for highly-available clustered storage. With Hadoop®, Ceph©, or BeeGFS® file system preinstalled, Unison achieves the right balance of metadata performance and data throughput.
The Unison Parallel File System Appliance, powered by Ceph, is configured to support object, block, and file storage, as well as OpenStack® and Hadoop® for maximum storage flexibility. Eliminate data silos and storage complexity with a solution built specifically for multi-tenancy and Cloud.
The Unison Parallel File System Appliance, powered by BeeGFS, is tailored for high-performance environments. Complex directories and small files are not a problem with Unison BeeGFS. The unique metadata architecture speeds up file creation, directory stat, and other functions that slow other platforms down.
The Unison Parallel File System Appliance, powered by Hadoop provides reliable, scalable, distributed computing. Map/Reduce distributes jobs for scale and takes advantage of data locality on each server, and data nodes are tuned and balanced with optimal memory to support the Hadoop JVM environment.
This white paper compares the performance of Unison, powered by Ceph, to that of the reference architecture as published by Red Hat. Unison demonstrated speeds of 1.8X for reads and 2.3X for writes versus their performance metrics.
Here are two case studies that illustrate how properly engineered storage systems can achieve industry-leading performance and open new vistas for bioinformatics researchers.
Increase capacity and performance easily and without downtime. Whether it is adding disks, clients, servers, controllers, or consoles, every element in Unison is built to scale. Unison eliminates data silos and storage complexity with a robust solution that lets you access massive data sets quickly and reliably.
OpenStack and Hadoop Integration
OpenStack Cloud, Swift, Glance, and Cinder integration is out-of-the-box easy with Unison and Ceph. Not only can it be the back-end of OpenStack, but also a stand-alone unified storage cluster. Unison with Ceph supports object, block, and file storage for maximum storage flexibility.
Built for Speed
Unison with BeeGFS is designed to be the fastest distributed file system available. This is storage packed with performance features such as distributed POSIX locking and native IB/RDMA, and a mature full-featured GUI to simplify management, monitoring, and expansion.
The Parallel File Solution distributes data for reliability and performance. With almost a decade of production use, these systems are rock solid.
Scalable Informatics purpose-built platforms are crafted for high performance and simplicity of deployment. From end-to-end, each solution is backed by Scalable Informatics’ expertise.
Distributed, parallel storage adds performance with each capacity node. The underlying hardware, networks, and Linux filesystem must also be tuned for the workloads, however.
Ideal for Cloud deployments with heavy parallel read requirements such as VM images, Ceph’s intelligent object model replicates data and balances workloads to deliver performance where you need it.
For bandwidth heavy applications, such as sequential reads and writes, BeeGFS/FhGFS benchmarks to near perfect scaling. In European testing on a 20-server cluster with 160 clients, BeeGFS achieved 94.7% and 94.1% of theoretical maximum.
Unlike other distributed file systems, BeeGFS metadata performance scales with servers. In the same configuration of 20 storage servers, 500,000 files could be created per second and over 1.38 million file stats were reached. A solution of 20 Unison BeeGFS appliances could create 1 billion files in approximately half an hour.
Scalable Informatics has extensively tested and tuned the Unison FastPath Appliance for balanced performance of network, compute, and storage. These systems are an expert coupling of hardware, system, and software to deliver the fastest platform and they are ready for expansion. Unison is tuned and optimized for storage from terabytes to petabytes.
Ceph, BeeGFS/FhGFS, HDFS, and GlusterFS have related architectures, but each filesystem manages metadata and data replication in a slightly different manner. All modern distributed filesystems are dependent upon each system’s local filesystem for performance, correctness, and fault tolerance. Each system is shipped configured by Scalable Informatics for your filesystem and storage configuration.
Scalable Informatics Unison appliances are preinstalled with metadata, data, and management servers to meet your scaling needs. Unison appliances can be scaled to mult-PB installations.
|Unison Flash||Unison Hybrid||Unison Disk|
|Form Factor||Rack-mount units in configurations of 2–10 servers per rack with management node|
|Capacity per Rack
• up to 10 machines per rack
|up to 9.7 PB||up to 6.0 PB||up to 6.0 PB|
• up to 81 racks per fully populated system
|up to 780 PB||up to 486 PB||up to 486 PB|
|Performance per Rack
• 10 machines per rack
|up to 300 GB/s R/W
up to 10M IOPs
( 30 GB/s • 1M+ IOPs per machine )
|up to 100 GB/s R/W
up to 5M IOPs
( 10 GB/s • 500k IOPs per machine )
|up to 70 GB/s R/W
up to 50k IOPs
( 7 GB/s • 5k IOPs per machine )
|Networking||Gigabit Ethernet management fabric • 40GbE, Infiniband FDR/ EDR, or 100GbE connectivity|
|Storage Options||Reliable Block, Object, Posix File System|
|Supported File Systems||Ceph • BeeGFS (FhGFS) • Hadoop • Lustre|
|Used For||Big Data • HPC • Cold/ Archive Storage • File System • Cloud|
|Support Options||Remote Support • Next Business Day Support • Depot Onsite|
Listen to an interview with Scalable Informatics CEO, Joe Landman, on choosing the right distributed storage platform.
Read the Transcript