For 21 years SGI has been virtualizing primary disk with tape using DMF, their storage tier virtualization software. This solution creates a seamless fabric across multiple tiers of storage, ranging from high-speed flash-based storage on the front end down through whatever disk performance is required. DMF marries it all with a tape back-end using an active file archive scheme to provide an “always available” data model. This enables the direct access that users demand and the economics that IT can afford.
But the data growth experienced by Big Data and ‘Big File’ industries (like Media and Entertainment and others that handle high resolution imagery), is creating the need for a a new storage solution, one between the performance of primary disk and the economics of tape. To bridge this gap in the storage infrastructure stack, SGI has added an object storage layer. Object storage provides the potential for enormous scale with the ability to maintain access performance, while containing cost.
Integrating the Scality object storage software with SGI’s Modular InfiniteStorage (MIS) platform creates a storage and compute system that’s extremely flexible and scalable. SGI’s MIS offering uses industry standard hardware to keep costs low and can even support third party hardware as well.
Scality’s peer-to-peer architecture uses a ring topology of servers, each with multiple storage nodes to distribute data, metadata and storage access. And within that ring topology, the system can support multiple hardware tiers, created by configuring the nodes with different storage and processing components, or by using different data protection options within the object storage software.
Then the system can move data onto the appropriate ring based on specific requirements for performance, data growth, data protection, etc. For example, a service provider may need a storage area with higher performance to support a large virtual environment and configure a ring with data replication and multiple data copies. Or, they can set up a ring using erasure coding to provide low cost, very high capacity storage area for content archiving. They can even set up a public cloud ring for DR purposes or additional cloud bursting capacity.
To create more flexibility and improve economics for large archives, SGI is developing a “MAID tier”, using technology from their acquisition of Copan a few years ago. MAID or Massive Array of Idle Disks puts disk drives into a lower power state (or powered off) when not in use. As discussed in the article “All MAIDS are not Created Equal” the SGI technology has some advantages over other MAID-based technologies which they say can reduce energy consumption by 85% compared with non-MAID systems. The system also runs continuous data integrity verification and regularly exercises each drive to maintain its reliability. This ‘zero watt disk’ tier brings more performance than linear tape can provide but at a much lower cost than even capacity disk arrays.
Data growth is pushing multi-tier archive systems to get bigger and bigger. But large file archives that use traditional disk storage architectures, even scale-out architectures, have some severe limitations when they’re asked to scale into the multiple PB range. Companies are forced to cluster multiple systems which increases management overhead and system complexity, as well as cost. And they can still fall short of capacity and performance requirements.
Object storage systems are a good technology fit for these kinds of file archives. With their ability to scale economically, with performance, object storage can enable companies to keep more data on a disk tier and still have the protection that’s required. SGI’s MIS system with Scality’s ring topology is taking that a step further by supporting multiple storage tiers within the object-based architecture. The result is a more flexible, more scalable, higher performing system that costs less, while providing a solution for both scale-out transactional storage and archive in one seamless fabric.