«The ClusterStor™ family of high performance scale-out storage for HPC and Big Data provides unmatched file system performance, optimized time to ...»
The Ultimate in Scale-Out Storage for HPC and Big Data
The ClusterStor™ family of high
performance scale-out storage for
HPC and Big Data provides unmatched
file system performance, optimized
time to productivity and the HPC
industry’s highest levels in integrated
solution reliability, availability and
Departmental Storage Scale
Supercomputing Storage Scale
ClusterStor™ integrated high performance scale-out storage for HPC and Big Data There is a feeling you get when you experience fine engineering. The feeling starts when you first lay eyes on the product and becomes more evident every time you use it. There are so many facets of your experience that tell you someone knows exactly what you need; innovative thought is evident behind every detail.
Xyratex has taken its 25 years experience in designing and delivering enterprise class storage, along with the Lustre® high performance file system, to create ClusterStor, an integrated family of high performance scaleout storage solutions for HPC and Big Data. Designed to provide unmatched file system performance and optimized time to productivity, ClusterStor provides the HPC industry’s highest levels of integration solution reliability, availability and serviceability.
This achievement is the direct result of a fresh ground-breaking design approach that meets the specific needs of high performance computing and its ever expanding scale.
The ClusterStor architecture is not merely an exercise of repackaging separate legacy storage hardware and file system software components.
ClusterStor is the result of state of the art optimization of the Lustre file system embedded within a modular scale-out hardware architecture to address a wide spectrum of HPC and Big Data application needs. ClusterStor offers solutions to best match your most challenging linear performance and capacity scalability requirements.
When you get up close, you can see that ClusterStor is not only innovative in its approach but also delivers a level of quality that translates into higher availability to meet your productivity requirements today and into the future.
Problem: Legacy storage metrics are not meaningful at the HPC and Big Data user level. Traditional Lustre file system solutions have typically used a mix of off-the-shelf components from different vendors leveraging legacy storage. Difficulties and roadblocks have been widely reported within the industry due to variances in vendor components interacting with separate file system servers and software packages, as well as the constant need to tune and retune file system parameters when attempting to add or even adjust file system processing capacity.
This traditional approach burdens HPC and Big Data environments with unacceptable program risk due to the excessive time, expense and non-productive work that constantly disrupts system productivity. What has previously been considered ‘maintenance’ is actually waste. Increasing HPC and Big Data productivity requirements demand that waste be eliminated.
The future of HPC and Big Data data storage requires a paradigm shift from legacy storage systems that focus on attributes like raw block storage capacity, controller cache size, or raw block data transfers.
Solution: HPC and Big Data user productivity is the only meaningful metric. File system storage productivity is best measured in sustained end to end file system throughput delivery, as well as reduced datacenter floor space utilization, cooling, power and administrative time. Unfortunately, traditional storage systems on their own are not able to address these user level attributes. A comprehensive solution paradigm is needed.
ClusterStor introduces a new standard in end-to-end file system performance and efficiency, at the user level.
Truly scaling file system performance efficiently means aggressively reducing the space, power, cooling and time consumed to achieve the HPC and Big Data performance and results users seek. Xyratex has proven that performance, efficiency and end-to-end solution management gains can be realized by consolidating system components closely together within an efficient, reliable and highly integrated scale-out data storage solution.
ClusterStor Architecture ClusterStor makes the solution paradigm clear. ClusterStor utilizes high performance, distributed computing data storage elements that have been designed and built from the enclosure and application platform all the way up to the operating system, file system and advanced User Interface (UI) management functions. Four high level solution components (storage hardware, operating system, file system and management application) have been synergistically bonded together to create the most reliable and best performing integrated high performance scale-out storage in the industry today.
Lustre’s future is directly tied to its quality and end user productivity success. This is why Xyratex makes significant contributions into Lustre testing, support and community involvement, and why we have honed the best attributes into our tightly optimized environment, eliminating variations and complexity.
The resilient ClusterStor operating environment introduces a scale-out storage and management architecture specifically designed to take full advantage of the performance and capacity scaling capabilities of the Lustre file system. ClusterStor doesn’t have to utilize special block-storage algorithms to predict how to provide optimal performance, the ClusterStor operating environment knows. In ClusterStor, the operating environment is integrated into distributed data storage elements called the Scalable Storage Unit.
Introducing the Scalable Storage Unit The Scalable Storage Unit (SSU) is at the heart of ClusterStor’s scale-out storage architecture. The ClusterStor SSU consolidates and integrates traditional block storage, network and file system components into a single building block that delivers a linear scaling unit in file system performance and data storage capacity. There is no need to guess at how to scale your Lustre file system: each unit is a balanced performance building block delivering a predictable level of performance and/or data storage capacity that scales linearly without storage or network degradation.
In a ClusterStor system, Scalable Storage Units (SSUs) act as independent I/O processing engines that work in parallel. Each SSU has its own set of resources that includes densely packed disk storage and dual embedded server modules, which also carry their own set of processing, memory and I/O network resources. This provides the most efficient use of each resource in a compact design that reduces the space required to deliver each gigabyte of file system performance.
ClusterStor 6000 - Delivers maximum performance scale for supercomputing environments To best address maximum file system through-put, ClusterStor 6000 delivers highest performance and data capacity density per SSU, achieving more concentrated performance per unit data storage capacity, while delivering excellent efficiency, reliability and ease of management.