Feature

Taking networked storage to the next level

Storage grids are here to take networked storage to the next level—closer to universal storage. By Akhtar Pasha

A large Indian business conglomerate is using elements of grid storage for seismic data processing in oil & gas exploration. They have deployed grid storage to hold down costs, streamline backups and ease storage administration. Then there are a handful of businesses, media houses and digital animation studios, that are evaluating elements of grid storage. None of them went shopping for grid storage per se. To them, grid storage is a useful technology for solving specific problems; it’s not a new product or technology in its own right.

When it comes to grid the most commonly talked about example that springs to mind is the electrical grid —a large, decentralised network with massive interconnectivity and coordinated management. Shailesh Agarwal, country manager-Storage, IBM India says, “If a subscriber wants an electrical connection he hooks his equipment up to an electrical grid. A grid station can supply electrical power to another grid station. It is possible as electricity is homogenous, which makes the grid a utility.” However he points out that large businesses have a wide variety of storage infrastructure (technologies) and are looking at storage virtualisation—they are looking at primarily heterogeneous network-based storage virtualisation products, which is not as homogenous as electrical grids are and hence cannot improve storage efficiency.

Grid vs. virtualisation

"Grid storage makes a lot of sense in India for certain applications as many Indian companies have distributed national networks linking regional offices, branches, remote locations, etc" - Ajaz Munsiff Director, Products and Solutions, EMC Asia Pacific

In grid storage you can add capacity as you need, or add a node to the grid and it self-configures and become part of an entire pool of storage reducing storage management costs. Virtualisation simply manages physical storage pools as one virtual pool, which is just one of the capabilities of grid storage.

Grid technology can provide many benefits when it comes to data storage. In terms of availability, for example, if one or more nodes fail, other nodes in the grid take over the task of delivering data to clients and applications. As resources are virtualised, tens to thousands of storage nodes act as a unified pool in a mesh like architecture. Each storage node can also monitor the health of all the other nodes. When one fails, any or all of the other storage nodes can help reconstruct lost data. Scalability also gets a boost. Every storage node deployed contains one or more terabytes of disk, as well as CPU processing power. This CPU processing power allows advanced data management algorithms to be applied to the data maintained by a particular storage node.

On the other hand virtualisation involves the consolidation of storage resources into a virtual storage pool so that when an application requires storage, the storage manager can allocate capacity from the pool. Since the resources allocated to an application can be increased or decreased at will, there is no need to over-allocate storage. Many analysts when they refer to storage virtualisation, they refer primarily to heterogeneous network-based storage virtualisation products. Analyst points out that with this approach, there are two inhibitors to improving storage efficiency. The first is that there’s limited support for functions that improve storage efficiency—most products go no further than providing simple pooling and copying functions.

The second inhibitor is that the complexity of deploying network-based storage virtualisation reduces the likelihood of it being used broadly by customers. That limitation, comes from constructing a solution out of extra layers of hardware and software, possibly from different vendors, each with different management and error diagnostics systems.

This complexity makes network-based storage virtualisation brittle and susceptible to significant management and general administrative overhead, as well as having an adverse effect on data availability and data integrity.

Storage virtualisation promotes greate rflexibility in provisioning storage as needed among multiple applications, providing an incremental improvement over provisioning storage based on estimates of each application’s requirements. However this benefit pales in comparison to the improvement in storage utilisation as a result of tiered storage—utilising virtualisation and replication to select the most appropriate storage for each application and stage in the data lifecycle.

Ajaz Munsiff, director, Products and Solutions, EMC Asia Pacific, says, “It is not easy to compare storage virtualisation with grid storage. It is like comparing traditional client-server computing (typically found in many data centres) with the grid-computing model. Both have their uses in different application scenarios. One can even argue that grid storage is yet another form of ‘virtualisation’.”

Grid elements in action

A storage grid is a meshed network in which no single centralised switch or hub controls routing. Grids offer almost unlimited scalability in size and performance because they aren’t constrained by the need for ever-larger central switches. Grid networks thus reduce component costs and produce a reliable and resilient structure. As enterprise storage environments get increasingly complex, organisations need the benefits that a storage grid delivers. For companies that make widespread use of very large IT systems, the storage grid holds the promise of making real-time scalability and near-infinite capacity always available. The first goal of a storage grid is if a qualified user needs additional capacity, it is made available with no appreciable management overhead. Optimised capacity also means that whenever storage capacity is added, throughput keeps pace. Analysts also agree that storage grids can reduce complexity of storage management, thereby simplifying administrative tasks and reducing operational overhead. This topography also improves utilisation management, by effectively increasing the utilisation level of each device and reducing the overall cost of infrastructure.

Munsiff adds “Grid storage makes a lot of sense in India for certain applications as many Indian companies have distributed national networks linking regional offices, branches, remote locations, etc.”

IT outsourcing will also be an avenue for grid storage adoption and growth. As IT services outsourcers look for ways to reduce their own capital and operational costs, grids may provide real cost relief while providing the service supplier with a flexible infrastructure to meet the demands of multiple clients or client workloads.

Think of a storage system that is built on the grid principles described above and implements all the popular NAS/SAN protocols in the same box. NetApp is perhaps only vendor currently heading in that direction because its technology supports both file and block level access and uses a distributed file systems, which providers a common virtualisation layer for all the storage in the grid. Soumitra Agarwal, marketing director, India, Network Appliance (NetApp) says, “Additionally we have Data ONTAP GX operating system, which is available on NetApp’s high-end FAS (fabric-attached storage) systems. It provides storage architecture the ability to more easily scale out horizontally but retain the notion of a single subsystem. So business can build a storage infrastructure to scale as broad as your compute infrastructure.” NetApp Scale-Out Storage System Data ONTAP GX systems are optimised to meet the demanding requirements of high-performance computing (HPC) applications running on large Linux and UNIX clusters. By combining multinode scaling and a clustered file system with NetApp FlexVol, Snapshot, mirroring, and RAID-DP technology, Data ONTAP GX systems provide performance, manageability and reliability.

“Data ONTAP GX systems are particularly well matched to the requirements of HPC markets. For applications such as seismic processing, movie creation, electronic design automation, genomics, and other data-intensive programs, there is a requirement to scale out storage capacity and performance to extreme levels, while reducing management complexity,” says Agarwal of NetApp. NetApp also has Data ONTAP 7G—OS, designed for storage grids and can pool multiple storage resources to make the capacity and power of the entire system available to all data sets.

Manoj Suvarna, country manager, HP StorageWorks Division- TSG, HP India says, “HP was first to deliver elements of grid storage in the marketplace in the form of ‘Proof Points’—such as HP ILM, RISS, XP1200 and Enterprise File Services (EFS) Clustered Gateway.”

HP has architected its storage grid from the ground up, based on the HP StorageWorks smart cell technology. The StorageWorks Grid is built from fundamental building blocks called Smart Cells. These are self-contained modular devices with integrated processing, storage and memory technologies. Smart Cells are organised into groups called Federations that interact and work together to deliver storage services. Each Smart Cell is given a specific trait that enables it to provide unique storage services, such as policy implementation, reporting, archiving, retrieval, auditing and anti-virus services. Addition, Smart Cells are dynamic and can be re configured to deliver a new, different service simply by loading new software. “The true power of this grid architecture is that each cell is aware of the other and can collaborate in order to distribute loads, increase overall performance through parallel processing and even cache I/Os meant for another personality. The entire grid system is presented to the administrator as a single unit and managed through various service interfaces,” says Suvarna.

Other building blocks that HP has for grid storage are HP StorageWorks Reference Information Storage Systems (RISS). For instance, there’s an active archiving solution that transforms data into useful information through application-aware, content-based indexing and powerful search tools. HP StorageWorks EFS Clustered Gateway—a NAS file server gateway that leverages existing and new SAN or iSCSI storage for either Linux or Windows. It can be configured into a cluster of EFS Clustered Gateway nodes. The clustering capability allows to scale file-serving performance in a near linear fashion as nodes are added to an existing cluster. In the event of a node failure, all file serving is failed-over to the remaining nodes within the cluster with full data and cache read/write coherency.

HP’s XP12000 disk array provides extreme reliability with no single-point-of-failure and redundant hot-swappable components. Multiple arrays and hundreds of terabytes can be managed from a single station by existing staff, with no additional resources. It supports multiple operating systems—HP-UX, Windows, Linux, Tru64, OVMS, NonStop, Solaris, AIX and NetWare.

Munsiff says, “The Centera storage platform which is actually based on grid storage architecture. EMC’s Centera storage platform deploys a unique, scalable, fault-tolerant RAIN (Independent Array of Independent Nodes) architecture for safely and securely storing data in a self-healing grid storage system. We have success in India for Centera and we have garnered many customers.” In addition, EMC’s Avamar data de-duplication software that transforms traditional back-up and restore also works on grid storage principles. It works with EMC and third party storage platforms.

Using Sun StorageTek 9900 Universal Volume Manager, customers can aggregate and manage multiple storage systems from different vendors under a single point of control. External storage volumes form a shared storage pool that can be used by multiple servers and multiple applications. Because different classes of storage systems can be virtualised as universal volumes, customers can create a tiered storage infrastructure and begin to implement data lifecycle management strategies where the cost of storing data is aligned with its value to the organisation as it changes over time.

Agarwal says, “IBM sees grid storage as only one of many capabilities needed to provide the many-to-many relationship customers want so that any application or user can have access to any computing or storage resources they needed—just like electrical grid.”

Agarwal of NetApp says, “A storage grid allows you to scale and create intelligent tiers of storage and allows you to have a single level of management to address multiple needs in a [storage] environment. All of these steps, what these are all leading to is a storage grid architecture being realised with real value.”

Questions to ask

With multiple approaches to grid storage, customers should ask vendors what they mean by grid storage specifically—questions such as what solution they have today, roadmap they have. Customers should ask what services, applications and storage protocols are supported and how grid architecture will be managed. Also customers should ask how to migrate from their current set-up to grid storage vendor. Be clear that grid storage vendors should have common management tool and support for both file and block-level data access and everyday benefits such as faster backups or restoration of data.

Additionally some industry pundits point out that grid storage should costs less 20 to 30 percent than the monolithic arrays—this price difference that combines with saving on storage and management, makes it worthwhile for some large businesses to buy grid storage.

Storage grids seem headed toward universal storage or global storage as some vendors put it that works much as an electrical grid does.