True to their name, enterprise arrays are massive in size and not for the budget-conscious; they are for organizations that support a large number of users and various types of high-performance data. Companies that can afford an enterprise array use it for their most vital data.
By consolidating storage to an enterprise array, users avoid the administrative tasks involved with direct-attached storage (DAS). With enterprise arrays, storage can be allocated to data that needs capacity, rather than having to assign capacity to each particular DAS system.
This tech roundup discusses how enterprise arrays are used, their evolution and what the competitive landscape looks like.
Enterprise-level arrays offer numerous processors, host ports, disk drives and large cache memory support. These arrays work by presenting logical unit numbers (LUNs) to attached servers. The servers may be directly attached to the arrays or attached through a switch. Security functions, such as LUN masking in the array or zoning in the switch, control which servers have access to which LUNs. Once a server has access to a LUN, it can then transfer data to the array over the network interface in use. Enterprise arrays typically have multiple interfaces.
Tony Asaro, senior analyst at Enterprise Strategy Group, Milford, Mass., says that one of the recent improvements of enterprise arrays is the ability to consolidate older enterprise arrays into one new system. "This allows customers to manage fewer systems and reduce software licenses and maintenance fees," Asaro says.
Key vendors and products
There are a number of enterprise array vendors in business and the competition is fierce. EMC Corp., Hitachi Data Systems (HDS), Hewlett-Packard Co., IBM Corp. and Sun Microsystems Inc. are some examples of key players in this game.
There are two prominent types of enterprise arrays: cache-centric storage systems and distributed-storage systems. Cache-centric arrays attach to both mainframe and open systems environments, while distributed systems attach to open systems only.
EMC has the Symmetrix DMX in the cache-centric space; IBM the DS8000 and DS6000; HDS offers the Lightning and TagmaStore; Sun the StorEdge 9900V line; and HP offers the XP12000 and 1024 and StorageTek the SVA. In terms of distributed-storage systems, EMC has the Clariion, while IBM has the DS4000; Thunder is HDS's array for distributed systems; Sun offers StorEdge 6920; HP the EVA and StorageTek the FlexLine.
"The battle for enterprise-class storage systems is a hard fought one," Asaro says. "EMC is struggling to stay on the top and HDS has done well gaining more customers directly and through its OEM partners HP and Sun.
"IBM has announced new products in this space that could give them a boost, but they should be doing better given who they are," Asaro adds.
Innovations and trends
Asaro says that he hasn't seen any enterprise array trends across vendors, and that each is taking its own path. HDS, IBM and a lesser known vendor named 3Pardata Inc. have had the most recent innovations.
HDS's TagmaStore Universal Storage Platform provides storage virtualization of heterogeneous storage systems, which according to Asaro, "allows for easier management, scalability and greater capacity utilization."
Asaro also highlighted IBM's ESS products –- the DS8000 and DS6000 –- both announced last year. The DS8000 supports logical partitions (LPARs), which allow users to create virtual storage systems within a single storage system. The DS6000 is a low-end enterprise array that provides high functionality for a lower cost. It is compatible with the DS8000, so users can do remote mirroring between the two systems.
3PARdata has been innovative in the area of thin provisioning, Asaro says. Thin provisioning allows users to create volumes of any size without actually purchasing or installing the entire amount of disk storage.
Randy Kerns, senior partner at the Evaluator Group, Greenwood Village, Colo., says that many storage arrays are starting to implement storage partitioning, where different images can be configured and allocated based on some controls.
"Partitioning addresses customer needs in moving data or consolidating storage systems, and adds a more granular level of service control," Kerns says.