Lustre Parallel Storage Solutions

As digital media and content continues to change our lifestyles and the way we work our need for increased storage and storage bandwidth continues to grow at a worrying rate. Storage needs have far exceeded expectations and clients are now looking for reliable storage solutions that offer genuine scalable performance without compromise for large bandwidth applications and high performance compute clusters.

Building upon our award winning range of Igloo storage solutions, our new range of Igloo "L series" parallel storage solutions provide genuine enterprise class storage that can be scaled to suit user needs whether it may be increased I/O performance, supplementary storage, further virtualised machines or improved availability and functionality.

Launched in partnership with Alces Software we assure our solutions deliver high throughput low latency performance to meet your requirements and budget.

Key Features

·         Dedicated High performance parallel storage appliances

·         Enhanced software tools deliver simplified management

·         Expand existing live systems on the fly

·         Support for over 10PB (Petabytes) of storage

·         Support for over 200GB/sec and over 25,000 clients

·         Add further storage nodes to suit your I/O performance and capacity needs

·         Upgrade & Extend existing Lustre systems

·         Discuss configuration and support services via remote access or on-site warranty

·         Discuss comprehensive training courses available including how to manage and extend file systems

·         Free software updates and annual onsite cluster health check

·         Onsite survey to measure current storage utilisation

·         Network and infrastructure assessment

·         Workload benchmarking to profile applications and data requirements

·         Perform data growth profiling to assess current and future needs

Specifications

Lustre Storage Solution	Raw Storage	Useable Storage	Write performance	Read performance
Small - single unit	10TB	7.2 (7336GB)	450 MB/sec	350 MB/sec
Small - two units	20TB	14.4 (14,672GB)	900 MB/sec	700 MB/sec
Medium - Single unit	22TB	14.4TB (14,672GB)	850 MB/sec	700 MB/sec
Medium - two units	44TB	28.8TB (29,344GB)	1700 MB/sec	1400 MB/sec
Large - single unit	32TB	21.6TB (22,008GB)	1350 MB/sec	1050 MB/sec
Large - two units	64TB	43.2TB (44,016GB)	2700 MB/sec	2100 MB/sec

Scalability through parallelism - what makes these solutions special?

Lustre works by considering your data files as a number of blocks of data where each block may be stored on one or more object-storage servers (OSS) with file metadata recorded by a dedicated metadata server (MDS). Compute nodes can mount the Lustre file system and use it like any other POSIX compliant file system under Linux providing compatibility with most Linux applications. As users write to the file system the compute nodes send blocks of data to the OSS's which write them to disk. By using multiple OSS's performance gains are observed as the I/O data is served by multiple storage devices in parallel. By adding more OSS's it is possible to achieve near linear scaling in aggregate storage performance over a HPC cluster.

Lustre supports a wide range of communication networks including Ethernet and Infiniband. Typically servers providing MSD and OSS services are connected to the cluster interconnect with all head and compute nodes. Files written by any node can be read and updated by any other node on the cluster similar to traditional NFS or CIFS storage solutions. To Expand the capacity or performance of the Lustre file system additional OSS machines may be dynamically added when required - the file system simply expands into the new space as it is attached.

·         Scale from TB to PB dynamically without downtime

·         Flexible deployment over Infiniband or Ethernet interconnects

·         Mix-and-match building blocks providing high capacity and performance

·         Single file system provides simple management and low cost-of-ownership

·         High density, power efficient storage servers

·         Certified and benchmarked configurations for speed of deployment

·         Ideal for video rendering, live broadcasts, media content storage and distribution

·         Parallel access to massive energy, manufacturing and imaging datasets

·         Snapshot long-running CFD and bioscience jobs to disk at gigabyte-per-second speeds

What is the Lustre Parallel File system?

The Lustre Parallel file system provides unprecedented storage performance to all nodes of a cluster for a wide range of data intensive HPC processing workloads. Whether you are performing a reservoir simulation across a multi-terabyte dataset, gene sequencing across millions of stored proteins or rendering the next Hollywood blockbuster, delivering data to your nodes is likely to cause a bottleneck.

The Lustre parallel file system provides a single file system namespace to every client node in your cluster thereby reducing the management overhead of multiple traditional file systems required to achieve scalability. With support for over 10PB (Petabytes) of storage, 200GB/sec aggregate performance and over 25,000 clients Lustre is the most scalable HPC file system currently available with 7 of the Top10 HPC systems worldwide using it in production.

                                                                                                                              By

                                                                                                                    K.  Pon abinaya

                                                                                                                      III  year CSE