When deployed in AI and Big Data applications or in high-performance computing (HPC), the performance of the implemented storage system lays a major role. Large amounts of unstructured data need to be stored and accessed quickly. It is also important that the storage used can grow with increasing requirements and that it makes data available again quickly in the event of a failure.
These are exactly the features offered by the IBM Elastic Storage System ESS 3000, a software-defined storage system for AI, Big Data and HPC workloads.
ESS 3000 combines the IBM Spectrum Scale parallel file system (formerly known as General Parallel File System - GPFS) for simplified data management with state-of-the-art NVMe storage technology for the highest performance.
1. High storage capacity
ESS 3000 is designed for IT applications that require the highest performance per rack. The storage appliance can accommodate a maximum of 24 x 15.4 terabyte (TB) NVMe drives in a 2U chassis. This equates to up to 370 terabytes (TB) of storage capacity per node, which can be increased to petabyte levels in a cluster.
2. Fast read and access speed
ESS 3000 not only offers high capacity, but also the necessary speed for modern machine learning workloads and traditional high-performance computing. The flash memory used is based on the latest NVMe technology, and the data throughput is 40 GB/s per 2U unit. Data is thus moved quickly to and from systems to keep pace with even the most powerful processors, whether IBM Power, x86 or GPUs. The latter can be fully utilized thanks to the high-performance memory layer, maximizing the performance of AI algorithms.
3. High data security
Software-defined "erasure coding" ensures that data is quickly recovered when needed. Spectrum scale erasure coding distributes data across available physical storage. This requires less space than traditional RAID, increasing effective capacity and data integrity. Data can be recovered in minutes instead of hours or days - with no business interruption.
4. Scalable without limit
IBM ESS 3000 adapts perfectly to increasing data management requirements in AI and Big Data applications. Storage capacity and performance can be easily increased by adding more ESS models. The IBM storage system can be operated in a cluster with other ESS 3000 units, but is also compatible with the ESS 5000 model.
5. Low TCO
IBM ESS 3000 features high storage density with up to 370 TB on 2U. The need for cooling and thus power is reduced, as is the required footprint. All this minimizes the environmental footprint in the data center and ensures a low total cost of ownership (TCO).
6. Ease of operation
With the containerized delivery model of the ESS 3000, the storage appliance can be commissioned quickly and easily. All required installation elements are efficiently packaged. Both setup and subsequent upgrades can be performed by in-house IT staff in a short time. As a result, the storage appliance is up and running productively after just a few hours.
Guardant Health uses Big Data and HPC platforms to transform massive amounts of genomic data into actionable insights for oncologists, researchers, and the biopharmaceutical industry. IBM Spectrum Scale's parallel file system provides high performance, while ESS systems deliver the data throughput our genomic pipelines require." Kumud Kalia, CIO at Guardant Health.
The test period runs until 30.08.2021. We will agree with all interested parties on an individually suitable time window for the test. Thus the test can be interrupted for a few days and resumed at a later date.