Delivering Results Built on Trust and Choice—for Big Data
Not all data is created equal. The value of data may not be known for years to come. A geographic survey from the past could yield information on prospective reservoirs. An abandoned jet engine design may provide useful insights in jet propulsion. The list of examples is endless.
What is common across nearly all data sets is that the underlying storage platform requirements change over time. Users typically need storage performance early in the data lifespan, within the first 90 days. As data sets age, organizations need an easy, efficient, and user-transparent way to move data to lower cost storage, such as tape. When storage systems need to be upgraded (typically every 2-7 years), organizations need an easy, non-disruptive path to new systems.
Cray now offers a way to do just this. The new system– Cray Tiered Adaptive Storage (TAS)–lets customers preserve data indefinitely, keep data continuously accessible to users and applications during migration, and upgrade the storage infrastructure for years to come, as needed.
In cases where customers need fast parallel storage, such as scratch, Cray offers Sonexion. Cray Sonexion reduces complexity, and scales performance and capacity together. Cray Sonexion consolidates Lustre in a compact, appliance-like form factor that interoperates with any popular Linux cluster.
Storage Solutions for x86 Linux
Got Linux? Think Cray for storage. All products and services provided by Cray’s data storage business for Big Data connect to Linux. Cray Cluster Connect (C3) for Lustre offers dependable, interoperable storage solutions for x86 Linux. Cray ships a Lustre Client for x86 Linux that comes packaged with C3. Cray optimizes Lustre across the entire data path from the Linux client down to the storage array. Customers choose the storage platform—DDN, NetApp E-Series, or Cray Sonexion—and Cray delivers and end-to-end storage solution that performs optimally.
Cray supports the entire solution from the Linux client down to the disks. What’s unique about C3 is the flexibility of the storage architecture. Characterizing how certain applications perform –and optimizing the entire I/O path from the client to disk—ensure the system scales optimally as needed. Cray’s expertise spans the entire stack—applications, compute, networking, and storage. Cray provides a single point of support, including all software and hardware, for multi-vendor storage solutions. Most storage vendors have developed expertise in a single area, such as block storage.
Cray holds the world record for delivering the world’s fastest in-production single Lustre file system at NCSA Blue Waters. Cray scales large sequential I/O performance from 5GB/s to 1TB/s in a single file system.
All of Cray’s storage solutions share a common benefit: delivering results based on customer requirements. Cray TAS is delivered pre-configured, deployment ready, and connects to industry-standard file sharing protocols like Lustre and NFS.
Understanding Parallel File Systems
For organizations investigating parallel file systems, Cray makes an ideal partner. There are many choices and decisions relating to software and hardware. Networking, file systems, and storage may be the most challenging for some organizations.
Where does Lustre fit? Should GPFS or NFS be deployed? Often, Lustre and other parallel file systems such as GPFS complement NAS and SAN deployments. Cray even has a way of virtualizing parallel file systems and NFS to maximize parallel access into Cray Supercomputers. This unique offering—Data Virtualization Services—comes included with the XC-line of Supercomputers running Cray Linux Environment (CLE).
Lustre is Cray’s native parallel file system of choice. Over two-thirds of the world’s fastest Supercomputers are powered by Lustre. As a co-founder of OpenSFS, a consortium dedicated to advancing open scalable file systems, Cray collaborates with industry partners and customers to advance Lustre.
Is Lustre ready for the enterprise and commercial HPC? Cray recently published a paper describing the decisions and considerations of using Lustre in commercial HPC and enterprises, where reliability is critical. Making the Business Case for Lustre is available for download on the Cray website.
Tiered Storage for Big Data and Large-scale Archiving
Cray TAS is ideal for customers requiring HSM-style simplicity using open source technologies and best of breed, multi-vendor storage technologies. Cray’s solution provides a flexible tiering model where customers can choose the media type—SSD, disk, and tape, in various combinations. Cray TAS abstracts file systems into a common storage so all data can be migrated bi-directionally between fast file systems like Lustre, across primary storage (often connected by NFS), and deep archives, usually tape-based.
In a strategic partnership with Versity, co-founded by Harriet Coverston, Cray has produced a complete tiered storage solution integrating SSD, disk, and tape. Cray builds on Versity’s open format HSM and storage virtualization software engine for Linux, and provides everything a customer needs to get up and going with an end-to-end archiving solution. This includes best practices, data migration services, and sample templates and policies. Customers classify data any number of ways, and maintain continuous access to data across its lifespan.
Use cases for TAS range from commercial enterprises supporting compliance and large-scale data archiving initiatives to tiered storage to digital libraries with massive archiving requirements. The need for Cray TAS could be driven by massive data growth (e.g., files and videos) or any number of company-specific data preservation requirements.
Cray develops all its solutions—from XC30 Supercomputers to TAS—with the future in mind. It’s essential to be able to upgrade systems over time to take advantage of the latest innovations.
As Seymour Cray once said, “the future is seldom the same as the past”.
Your Trusted Expert is Cray.