Clustered network storage: part one; Smarter, faster, cheaper and easier.

There is a quiet revolution taking place with network storage, and the early adopters are reaping the rewards. A new wave of storage systems that support clustered architectures are available today for SAN, NAS and Object-based storage. Clustered network storage is a major advancement that can significantly reduce cost (hardware, software) and greatly simplify the management of storage.

Traditional storage systems typically have two controllers that are clustered together creating an active-active pair that appears as a single logical storage system aggregating processing power, bandwidth and memory. Storage systems that support active-active configurations have become the predominate architecture used by all of the leading storage vendors including EMC, HDS, IBM, NetApp, etc. These solutions are limited by the number of host ports, cache memory and storage they support. They range from 2 to 8 FC and/or Ethernet ports, 2 to 4 processors, from 1 GB to 8 GB of cache memory, and from 500 GB to 60 TB of storage. These storage systems are typically classified as mid-range or modular and are experiencing the greatest market growth in network storage.

A high-end mid-range storage system that supports 8 host ports may not provide enough bandwidth for a large server environment. Additionally, since mid-range storage systems typically support only two processors they may not have adequate compute power to handle a large number of transactions, which could impact application performance and ultimately user/customer satisfaction. This is also true of cache memory, which has been proven to greatly improve performance for sequential data streams. If customers run into performance and scalability bottlenecks using traditional active-active storage systems, their only recourse may be to buy another system.

There are a number of new storage systems in the market that have solved this scalability problem by supporting anywhere from 3-way to 32-way active nodes in a cluster. And regardless of the size of the cluster it is managed as a single system. The implications of this are enormous. Customers can scale these systems from a two node controller configuration and just keep adding nodes as the demands of their environment change and still only have a single system to support.

The economics of a clustered network storage system are compelling with the potential of significantly reducing capital and operational costs. With a clustered network storage system the customer only has to add another controller node, which costs far less than acquiring a whole new system, its associated software and additional maintenance charges.

Customers can scale and grow network clustered storage systems from small, medium to large configurations just by adding more controller nodes to the system. Every node added to the cluster increases the number of processors, host ports, cache memory and the supported capacity. Most of the leading storage vendors have Enterprise-class, mid-range and low-end systems that are incompatible with one another using different software and hardware. This increases the complexity and cost of migrating up (or down) if this is required. The new way offered by clustered network storage systems is the ability to grow from low-to-mid-to-high using the same system.

There are SAN, NAS, and Object-based solutions that have clustered network storage architectures. The SAN solutions include the Compellent Storage Center, Digi-Data Xtreme, EqualLogic PeerArray, Intransa IP-3000, 5000 and 7000 series. LeftHand Networks, Sun 6920, and XIOtech Magnitude 3D and 3D Edge. The NAS systems include EMC NS and CNS, Exanet Exastore, Isilon, OnStor SAN Filer, and Panasas ActiveScale. Object-based solutions include Archivas ArC, EMC Centera, HP RISS, and Permabit Permeon. Each of these solutions has a wide range of features, capabilities and architectural differences that require a deeper analysis than will be covered in this article. It is important to point out that all of the solutions mentioned above are being used by customers today that chose these products over traditional storage systems.

There are counter arguments that can be made on the side of the traditional 2-way clustered architectures. For example, FC host ports will soon support 4 GB of bandwidth allowing for a greater ratio of host servers to storage ports. Additionally, customers are consolidating servers so the number of FC host ports needed on a storage system may not be an issue. However, the cache memory and processing power are still limited and not scalable and at some point may be a bottleneck. The most valid argument that vendors can make about their 2-way products is that their customers are not having performance and scalability problems and it is a non-issue. The most invalid argument is that clustered network storage solutions that support more than 2-way architectures can't work because inter-communication between nodes hinders scalability. ESG Lab has tested a number of these solutions including Compellent, EMC NS/CNS, EqualLogic, Exanet, Intransa, Isilon, OnStor, and XIOtech. We also spoke to dozens of very happy customers using these various products in their production environments. These clustered network storage systems work and work very well.

The name of the game is ease of management. Customers can grow as needed (processing power, bandwidth, cache memory and capacity) and still have a single logical storage system to manage. The goal is to be able to support 100 TB, 500 TB and even 1 PB of storage as easily as you manage 1 TB of storage. Just as important is reducing the cost of acquiring new storage by adding on to your existing storage system versus buying new ones that come at a premium even for just a small amount of storage. There are customers that are achieving these goals today. Are you one of them?

Tony Asaro is a senior analyst for Enterprise Strategy Group (Milford, MA)

www.enterprisestrategygroup.com

Opening shots in continuing stories ...