Apr 7, 2014
Traditionally, storage has been defined as either midrange or enterprise, based on the assumption that midrange or smaller customers did not need the performance and availability of larger enterprise customers. As a result, midrange storage could be built on a simpler, lower cost, dual controller architecture, where one controller supported the other controller in an active/passive mode. In other words, the data that was being stored in one controller cache was copied into the other controller cache so that if the primary controller failed the data in its cache could be protected and operations could continue through the other controller. This would ensure no data loss, but could impact performance since the surviving controller would have to support the entire workload. There is also the danger of data loss if the surviving controller also had a failure, so in most cases, down time is scheduled to restore dual controller operations. Dual controller storage arrays are fine if you can afford scheduled down time.
Enterprise storage arrays eliminate the need for scheduled down time by providing 8 or more controllers that can share a global cache that is write protected. More engineering is required for this type of architecture, to provide shared connectivity, and mapping of the cache to the different port processors. The additional controllers enable greater throughput, scalability and availability than dual controller storage arrays, but are more costly. The additional controllers in these types of arrays also enable them to provide more functions like replication without impact to I/O performance.
The classification of storage architectures by the size of the customer is no longer valid in today’s 7 x 24 business environment. While smaller customers may not need the same throughput and scalability as enterprise customers, they do need the same storage availability as enterprise customers. It is also no longer true that enterprise customers need enterprise storage for all their storage requirements. Enterprise customers would like to store the bulk of their less active data on lower cost storage arrays in order to lower their costs. Both midrange and enterprise customers need high availability, feature rich storage as well as large capacity lower cost storage. The difference between midrange and enterprise customers is just a matter of throughput, scale and budget. The business requirements and not the size of the customer should define the storage infrastructure. Classifying storage arrays as either midrange or enterprise is no longer a valid way of defining storage requirements. The business requirements should define the storage infrastructure without regard to the size of the business.
Hitachi has designed two types of storage arrays, one type being the Hitachi Unified Storage 100 (HUS) family with dual active controllers, that provides the best storage economics for high performance file and block platforms like Exchange, Oracle, or webservers. HUS 100 has an automated load balance feature that enables ports on either controller to access data in the primary controller cache and provide continuous operations during scheduled or unscheduled maintenance.
The other type of array has an internal switch architecture with global cache, which can scale up and out to meet the demands of large virtual servers and of multiple bare metal servers. This type of array also supports built-in virtualization of external storage systems, which are simply attached to the FC ports of this array and presented through its global cache as though they are part of the array. In this way the external storage inherits all of the enterprise capabilities of the virtualization platform that it is attached to. Data can be moved automatically to the right tier of internal or external storage based on business requirements. This type of Hitachi storage array comes in two forms; an HUS VM, which is priced and packaged for the enterprise that does not expect to scale beyond a certain level and the VSP for the enterprise, which expects to scale almost indefinitely. While there are other ways to provide virtualization through an appliance that sits in front of the storage systems, appliances introduce an unnecessary level of additional network connectivity and management costs without delivering the full capabilities of an enterprise storage system.
A midrange customer can start with HUS VM, enjoy all the enterprise functionality available in the larger VSP, including virtualization and automated tiering between internal and external storage. The cost reduction in the HUS VM comes from large-scale integration of four types of Hitachi processors into one, the latest Intel multi-core technology, the exclusion of mainframe support, and scalability that is limited to most midrange or standalone environments. HUS VM is limited in scalability to 16 processor cores and 16 director cores, 48 FC paths, 256Gb of cache and 1152 disk drives, which should be able to cover most midrange requirements for file and block storage. As data becomes less active it can be automatically tiered to lower cost HUS or other “midrange” storage systems that are virtualized behind HUS VM. An enterprise customer can start with a VSP containing 8 processor cores and 4 director cores, 16 FC ports, 64 GB of cache, virtualize existing storage systems, and immediately upgrade legacy storage systems with the latest enterprise capabilities of VSP. VSP can to scale to 32 processor cores and 64 director cores, 192 FC ports, 1 TB of cache, 2048 internal disks, and 255 PB of internal and external capacity. Large enterprises could also consider HUS VM for stand alone or distributed business requirements.
When either midrange or enterprise customers have business requirements for single use platforms like exchange or web servers that require high response time performance, and high availability with no data loss, they should consider the HUS 100 family of systems. HUS 100 with large capacity disk drives are also ideal for low cost tier 2 or tier 3 storage sitting behind HUS VM or VSP.
Storage should no longer be classified as midrange or enterprise. Midrange and enterprise customers have the same business requirements for different types of storage. The differences are primarily in cost and scalability and the value that business place on the data. Some small companies whose entire business depends on the availability of their data may decide to go only with enterprise storage systems like HUS VM or even VSP.