Morphing from workaday drudge to the belle of the ball, storage is the Cinderella of IT. Its traditional, and often ancillary, role as a safe and reliable repository for data hasn't changed but the amount and variety of information that companies manage have increased dramatically. There's no relief in sight to this data glut, which was kick-started by companies that moved their businesses to the Internet and embraced multimedia formats, such as graphics and movies. "Storage is a bad word for storage because it is access and movability that is driving sales," says Will Brenyam, IT manager for the Special Broadcasting Service (SBS).
Networked storage promises both unrestricted capacity and a flexible architecture to better manage varying business requirements. But its complex implementation poses new challenges for IT departments that must master unfamiliar emerging technologies. Nevertheless, networked storage offers obvious advantages over DAS (direct-attached storage), such as hard disk drives or RAID arrays. At Hewlett-Packard's recent ENSA (Enterprise Networked Storage Architecture) conference, Neal Clapper, vice president of online storage solutions, said that while DAS storage makes up a large percentage of storage today, approximately 50 per cent of it goes unutilised. "Storage is possibly the most mismanaged asset in an organisation," Clapper says.
Even more tantalising is the unprecedented freedom a storage network gives to run business-protecting tasks, such as data backups and volume replications, with little or no involvement from an application server. This independence translates into storage managed as a resource in its own right rather than as an extension of servers and applications.
Clearly, storage is one of IT's costliest, most pressing problems. The usual approach to deploying and managing storage devices has become impractical and, in most cases, unsustainable. Simply piling more capacity on top of the old is no longer the answer - different types of data require different storage methods tailored to the unique business requirements of each company. Networked storage is invariably part of the solution.
The majority of businesses have already implemented, or are considering a networked storage solution and have adopted multiple storage protocols. In a recent survey conducted by InfoWorld in the US, 55 per cent of respondents indicated that they had deployed a file-based NAS solution, whereas 25 per cent have deployed the block-oriented SAN. Still, DAS remains a fixture with 52 per cent of readers also reporting that they continue to use DAS. Meanwhile, Jim Bernhard, director of strategic development for Adaptec's storage networking group, says SANs (storage areas networks) are increasingly being used as a way of solving manageability issues associated with large and ever increasing amounts of disk storage. "The SAN approach offers many storage management advantages, including the ability to share a pool of storage, easily grow the size of the storage pool, balance the load to each physical device, share backup devices, and manage computing resources separately from storage resources," says Bernhard.
The first step is to match data requirements to storage solutions, which requires an accurate assessment of the data and an open-minded understanding of its business relevance. The storage market now offers an unprecedented set of tools and novel technologies that can tame the stickiest problems. For example, although SCSI remains the dominant interface for storage devices, complementary transport protocols such as iSCSI (Internet SCSI) are challenging the well-established but budget-unfriendly FC (Fibre Channel) to allocate storage devices on the network.
The promise is great, but networked storage isn't cheap. How much companies spend on networked storage is determined by the extent of their storage problems and their budget capacity. For many IT leaders, this is uncharted territory - a great many don't know how much they will spend during the next 12 months on storage hardware, software, and services. Meanwhile, the indicators from Gartner, and even the projections of storage vendors such as HP and EMC, suggest the coming years are going to be lucrative. Gartner predicts 11 per cent compound annual growth rate over the next five years and HP expects to earn $US84 million from its storage group covering Australia, New Zealand and Fiji.
On the integration front, a backlash seems to be unfolding against the overwhelming desire to simplify. "Management keeps telling us that we've got to demystify and simplify IT and the reply is, well, no we don't, because it's not that simple'," says SBS's Brenyam. Similarly, Andrew McLorinan, newly appointed general manager of Hallmark Computers, warns that while "storage area networks have come of age, only a special few resellers can identify the opportunity and solution sell adequately".
"The centralisation of SAN resources is as much a question of organisational change as of technology, so from a sales perspective it is more complex and challenging," says McLorinan. "Nevertheless, the business case is often so compelling that the SAN sale will coincide with significant server and desktop business."
Respondents to a recent InfoWorld survey report a staggering combined total of 21,500TB of data:
- 45 per cent of those terabytes are allocated to host databases- 21 per cent are dedicated to user files- 15 per cent are used for e-mail messages and attachments- 6 per cent go to Web content- 13 per cent go to other applications
Ian Selway, network storage solutions guru for Hewlett-Packard Australia, says that while iSCSI has a lot of potential, it is unlikely to emerge as a mainstream interconnect technology immediately due to standards issues (the iSCSI standard isn't yet fully defined, ratified and adopted) and a lack of iSCSI adapter cards. "Most importantly, to function adequately iSCSI requires 10GB Ethernet and this isn't likely [as a cost-effective solution] until at least the end of 2003," Selway says.
"The other major question mark surrounds the development of mechanisms to undertake the processing, thereby relieving system CPUs of this function. These TCP/IP offload engines (TOE) are being worked on but to date the cost is not economical and the lack of bandwidth and chip performance to handle this TOE functionality is stalling iSCSI development."
Fibre Channel is, and for the foreseeable future will remain, the primary interconnect architecture linking applications servers and storage devices. Over time, however, IP will become an increasingly important transport mechanism seamlessly working with Fibre Channel to provide greater scalability and extend the storage infrastructure over longer distances and at reduced costs.
FCIP is a protocol that creates a Fibre Channel tunnel in an IP network to connect two SAN islands together. FCIP is the preferred choice for data replication and interconnecting SAN islands because it preserves full compatibility with Fibre Channel SANs. It uses a protocol converter to encapsulate the Fibre Channel data stream in an IP tunnel and routes the data on an IP network. When the data reaches the remote site, a protocol converter extracts the Fibre Channel data stream from the IP tunnel and routes it on the remote Fibre Channel SAN. The IP tunnel acts as a dedicated link to transmit the Fibre Channel data stream over the IP network, while maintaining full compatibility with the Fibre Channel SAN.iSCSI is a protocol that allows a server to access storage through an Ethernet network adapter. Performance problems and embryonic standards make it unacceptable for most applications today. SANs would be built with standard Ethernet IP switches and integrated into the corporate LAN backbone. Servers would connect to the Ethernet switch with IP network adapters, and the storage system would have IP host ports that connect to the Ethernet switch. However, the industry must address several problems before IP becomes a viable alternative to Fibre Channel, which means it is unlikely to displace Fibre Channel in the next two to three years.iFCP is a protocol that translates communication from Fibre Channel devices, like server HBAs and storage systems, into an IP protocol compatible with a standard Gigabit Ethernet network. The goal is to replace some or all of the Fibre Channel fabric with a Gigabit Ethernet fabric, while preserving the Fibre Channel interface on the server and storage system. This could enable SANs to extend across IP metropolitan-area and wide-area networks. Because iFCP supports Fibre Channel devices it has better potential for compatibility with some of the advanced storage solutions than iSCSI.
FCIP accomplishes the same goal as iFCP for connecting two SAN islands with a point-to-point tunnel across a metropolitan or wide-area network. However, because iFCP converts Fibre Channel into a general-purpose IP protocol, the SAN becomes part of the native IP fabric, able to connect many SAN islands instead of the point-to-point connection supported by FCIP.
The most interesting part of attending HP's recent Enterprise Networked Storage Architecture conference in Queensland was listening to users' tales. Many users with wildly different backgrounds and technical requirements described how and why they moved to a new storage solution. But whatever their situation, the successful users shared several attributes.
The first and most important was an open mind. These IT managers did not have a preconceived objective to stick to, such as "Give me SAN or give me death!" Rather, they analysed the major problems at hand then started exploring technical solutions. "Decision-makers are looking at the larger storage networking market rather than the individual SAN and NAS components," says Harry Christian of Network Appliance. "As the purchasing decision moves up in the organisation, specific technologies matter less - what is important is that a solution can meet the business requirements."
Secondly, senior management was involved in the companies' storage migrations. CIOs reported that storage backup has become more important to their managers, a clear sign of a radical change in attitude (despite popular opinion, this was not sparked by September 11). Adjusting a backup strategy can be astronomically expensive and to grant funding, senior managers must understand the benefits of your storage project. More importantly, they have to understand the cost of inaction - when it comes to storage, doing nothing can become more expensive than deploying a new solution. Mind you, customers report that this meeting of minds is no menial task. "There is no common conversation between management and IT," says SBS's Will Brenyam. However, IT managers are finding innovative ways around these hurdles. For example, Mark Charmichael of Ernst & Young says he approaches senior management as if they were clients to provide a familiar framework in which to operate.Thirdly, the IT leaders were realistic. They all confessed that their solutions were probably not the best possible choice per se, but they were the best for their company's objectives. The majority emphasised efficiency, leveraging existing resources and controlling costs. "There is higher customer demand for low total cost of ownership and the need for scalable, flexible storage resources," says Christian. "Users are also confronted with the need to deliver heterogeneous server platform support to reduce management overhead in the face of increasingly complex and growing storage deployments and to ensure data protection and rapid recovery in an environment lacking a window for backup. Also driving demand for flexibility is the movement towards storage on demand or just-in-time storage purchasing."Successful IT managers leveraged the storage vendor's knowledge to assist with their project, saving significant consulting fees and, possibly more important, attaining a commitment from a major vendor to the success of their project.
It takes years to develop a relationship with a vendor and the best time to harvest the fruit of that rapport is when engaging in a challenging new endeavour. CIOs also advise others in their shoes to think long term but act short term. When companies get started, replacing the technology behind their storage infrastructure can easily take years to complete. Failing to appreciate that simple fact is a recipe for disaster.