Grid expert Mark Linesch discusses how standards work on storage grids could finally get multivendor storage systems to interoperate.
Other stories on this topic NetApp’s bid for grid
Yes, but it's also the notion of location. In a gridlike structure, storage can be in one place or in many places, [and] it can be from one vendor or many vendors. When you talk about shared storage and clustered file systems, you are looking at a single-vendor solution.
What types of enterprises are deploying storage grids today?
Companies that are managing large amounts of customer data, for example, financial institutions, banks, brokerages and retailers. Grids started in the high-performance-computing area, primarily with engineers and scientists, then moved toward very-high-performance line-of-business applications in finance, pharma, energy and manufacturing.
What work is the OGF doing toward managing grids, file systems and virtualization?
We are working on a variety of standards for the compute, network and storage infrastructure, all the way from describing jobs to being able to move and manage data. We have working groups that are defining a grid file system and developing a naming scheme, which provides a way to manipulate storage without having to worry about the physical location of the storage. We're also doing work around a storage resource manager that can optimize distributed data within a grid.
OGF's specifications aren't done in a vacuum: You have the DMTF [Distributed Management Task Force], which is working on a Common Information Model [CIM].You have the Storage Networking Industry Association [SNIA] with its Storage Management Initiative Specification, which is aligned with CIM and is very important to this generation of storage-area networks and will be very important to generations to come that are more gridlike. SNIA has a grid initiative, and OGF has a Storage Networking Community group. Together, we look at the possibilities and try to align our initiatives with each other but also to develop new best practices, new knowledge and specifications.
By using the OGF's Global File System or Grid Storage Management standards, will vendors be able to make their proprietary software interoperable?
Yes, vendors will be able to define interoperability points using some of these standards. You will be able to get to a higher level of abstraction in a grid file system, so that it can comprehend the different file systems out there in a grid environment. That's happening right now: You see a lot of innovation in the storage area with start-ups that are doing storage-gridlike things, and you see standards organizations like the DMTF and SNIA working to figure out where the appropriate opportunities for standardization are. Right now, we're between innovation and standardization.
To use products from your former employer as an example, in the HP StorageWorks Grid, the Reference Information Storage System uses the ext3 file system, while the StorageWorks Scalable File Share uses the Lustre file system. In a grid, how do you accommodate these different file systems?
You have a specification that has a generalized naming scheme, and you abstract those proprietary file systems up to a higher level so that the data in the grid can be in any format; it might be stored in any type of storage system, and the standards mechanism describes and organizes the file-based data. Our Grid File System Working Group is providing a specification for a grid-file-system directory service and a generalized architecture for a grid-file-system service. Then those unique file systems can plug in or talk to the service and tap into the generalized naming or directory service. [Note: Working group members include IBM, San Diego State University and the National Institute of Advanced Industrial Science and Technology.]
When we talk about grids, we see storage grids and compute grids. When do you see those grids converging?
In many cases they are converged already. If you think about a gridlike fabric, clearly there are instances of that from a storage and server perspective. It is the application, however, that drives grid use.
Storage grid standards and specificationsStandards from a variety of organizations are creating and converging storage and computing grids
Standard or specification
What it recommends
From the Open Grid Forum
Grid File System
Standard interface and architecture for a logical file system
IBM, San Diego State University
An information document outlining requirements and architectural considerations was published in January 2006. An associated naming specification (Resource Naming Service) is expected to be published during 2007.
Grid Storage Management
A storage-resource management system to provide dynamic space allocation and file management
Lawrence Berkeley National Laboratory
The specification is expected to go into the working-group last-call stage in May 2007.
Open Grid Services Architecture
A blueprint for heterogeneous, interoperable grid and storage-grid computing
Published Dec. 2005 to May 2006.
Grid Scheduling Architecture
An architecture for scheduling jobs on the grid
Forschungszentrum Jülich, Universitat Dortmund
Work is in the early stages. The group published use cases in March 2006.
From the Storage Networking Industry Association
Storage Management Initiative-Specification
Heterogeneous management of storage devices from a single console
EMC, Hitachi, HP, Sun, Symantec
Version 1.1 of the specification was approved in 2005.
From the Distributed Management Task Force
Common Information Management
A common definition of management information for systems, networks, applications and services including grids.
Cisco, EMC, HP, IBM, Intel, Microsoft, Sun, Symantec
CIM Infrastructure Specification was released in 2005. CIM Schema draft was released in December 2006.
< Previous story: Virtualization tames the storage beast for vet school |