What are you looking for ?
Infinidat
Articles_top

Clustering of infinite io NSC-055s Network-Based Metadata Accelerator

"Performance exceeds fastest storage systems" to reduce workflow processing times.

infinite io, Inc., announced the clustering of its NSC-055s network-based ‘Metadata Accelerator’ for scale and system fault-tolerance.


Metadata is data describing data. By looking at metadata, applications discover the status and state of data: things like last access time, file size, when was the last time a file was saved, or is a file open or closed. Storage systems are optimized to move data, not to respond to metadata requests, and their performance quickly degrades as metadata requests increase.

With the clustered accelerators, enterprises will see an improvement in storage system performance as the NSC-055s off-loads metadata processing from installed storage. Metadata processing is put directly in the network, freeing storage systems from the performance-killing burden of serving up metadata requests.

It is a well-known fact that in the typical storage workflow, most of the requests made to storage systems are metadata requests,” said Mark Cree, CEO, infinite io. “Once installed, we’ve seen the NSC-055s offload as much as 90% of all requests going to storage, and our customers have seen dramatic reductions in workflow processing and response times.

Installs like network switch and makes metadata acceleration transparent
Clustered NSC-055s install like a network switch. It does not require applications to be re-provisioned or storage to be virtualized with new file paths. Installing transparently in front of existing storage, the NSC-055s intercepts and responds to metadata requests.

The NSC-055s is a no brainer for simulation, design, rendering, software build, and EDA applications where it can deliver performance numbers that no storage system alone can match,” said Steve Duplessie, founding analyst, Enterprise Strategy Group. “infinite io’s transparent approach makes this a simple plug-and-play performance enhancement.

Performance no storage system can match
A three-node cluster of the NSC-055s can serve up to 3,000,000 metadata requests per second with an average latency of 65μs. For a comparison, the fastest, all-flash arrays typically respond to metadata requests with an average latency of 500 -1,000μs – over ten times slower.

With infinite io installed we saw a significant reduction in workflow processing times,” said Rhian Resnick, assistant director, middleware and HPC, Florida Atlantic University.

Clustered accelerators contain copies of all the metadata stored in the attached storage. The copies are kept current in real-time using deep packet inspection.

Increased scale and fault-tolerance
The clustered accelerators provide expanded connectivity and enhanced fault-tolerance. Multiple accelerators can be clustered to form a single logical unit, allowing customers to grow network connections as needed by simply adding accelerators. Clusters start with three accelerators and twenty-four 10Gb/s port pairs. Clustering accelerators also provides a new level of redundancy that meets or exceeds traditional storage systems and can survive accelerator, network, and storage system failures.

Clustered NSC-055s are available for shipment. NSC-055s clusters are software upgradeable, while in production, to network-based storage controllers which provide additional features to transparently migrate inactive data off installed primary storage to a private or public cloud.

Articles_bottom
AIC
ATTO
OPEN-E