Enterprise Storage: Storage Is Like a Box of Chocolates
February is the month for confectionaries and chocolateers, who annually sell several billions of dollars of candy as part of the observance of Valentine's Day, a celebration held over from the Lupercian fertility festival of Ancient Rome.
Pope Gelasius tried to take the fun out of the Roman holiday in the year 496 by associating it with martyred Saint Valentine, who was beheaded for facilitating illegal marriages between Christians. Despite the good intentions of the Pope, however, the combined attraction of sugar, romantic lingerie and funny greeting cards proved too much even for the spiritually inclined. Today, on Valentine's Day, one is still likely to see both gentlemen and philanderers exiting gift shops with brightly-colored, heart-shaped boxes tucked under their arms for use as barter and bribes.
In case you've never seen one, the modern Valentine's candy assortment is largely comprised of sweet nut, fruit or cream concoctions wrapped in envelopes of chocolate, continuing a tradition that dates back to the perfection of sweet chocolate by the Swiss in the 1700s - some 200 years after the "discovery" of cacao beans by Christopher Columbus in the New World. However, as has been observed by the British comedy troupe Monty Python, some vendors have introduced of late some rather unconventional variants on the traditional "chocky" - such as crunchy frog, cockroach cluster, ram's bladder cup and a surprise candy that fires sharpened metallic bolts through the cheeks when eaten!
Which is kind of a long way of getting around to the point of this month's column. Storage, to paraphrase the great literary philosopher, Forrest Gump, is like a box of chocolates. You're never sure what you're going to get. Here, then, are my storage predictions for 2001.
Disk capacity/price curves will reach their peak. With current disk storage technology reaching its 150 GB per square inch areal density limit (as dictated by the superparamagnetic effect), 2001-2002 will see the end of the 120 percent annual capacity improvements we have witnessed since the late 1990s and probably, also, the 50 percent-per-year reduction in disk drive price. This transition will probably not be felt by consumers until 2004, but vendors will need to begin preparing now to address a single issue: How they will be able to perpetuate the increasing capacity/decreasing cost value proposition to their consumers? The answer will be networked storage: Virtual volumes will need to deliver what physical devices no longer can if storage is to keep pace. Expect greater attention by vendors to smoothing the way for heterogeneous, interoperable and scalable storage architectures beginning in 2001.
IP-based storage arrives. In line with the above, expect the arrival of IP-based storage networking products in 2001. Most will embody a pre-standard version of the new "iSCSI" (SCSI over IP) protocol being developed within the IP Storage Working Group of the IETF. In addition to the groundbreaking EtherStorage product family from Adaptec, look for iSCSI host bus adapter (HBA) products to be announced by one (or more) of the most vocally pro-Fibre Channel, anti-iSCSI, SAN adapter vendors in a startling reversal of the company's earlier caustic rhetoric. Cisco Systems' entry into the SAN space is a powerful persuader.
SANs will change. What we currently call a SAN is a primordial storage network - a typically homogeneous server-storage interconnect fabric masquerading as a network. Expect SANs to change in 2001: from island architectures to federated architectures to universal architectures. SAN "operating system" issues will come to replace network plumbing issues - ushering in a new series of inter-vendor bickering. Various SAN OS approaches will be embraced by different server operating system vendors. In the process, new obstacles will be created to hamper the realization of a single, universally-accessible storage pool.
Expect a challenge to traditional SAN from new SAN/NAS hybrids. EMC's acquisition of CrosStor Software and Network Appliance's pursuit of the Direct Access File System protocol are harbingers of a new generation of storage products that will deliver the benefits of high-speed, block-level access found in large arrays, the scalability expected of SAN storage, and the ease of implementation and single point of management offered by NAS. SAN/NAS hybrids will effectively challenge the original vision of SAN by delivering scalable, high-speed, well-managed storage without the kluge of do-it-yourself wiring and switching, combined with limited utility, third-party, SAN management tools.
Expect storage virtualization to become a network service. Complex, microcode-heavy array controllers are on their way out. They are obstacles to the realization of virtual, shareable storage volumes. Moreover, they add to the cost of storage, which is an inhibitor of product sales in the middle tier of companies - the only segment of the market showing significant growth. Host-based virtualization software is also a non-player: too much to maintain on too many host systems. Host-based virtualization is a driver of storage management costs. Instead of the controller or host-based software providing the point of virtualization in a storage network, SAN intelligence will shift to the network itself. StorageTek, DataCore, NetConvergence and Vicom are all doing interesting work in this space.
Traditionalists versus innovators. While the folks at EMC are fond of saying that the world of the future will have only two types of companies - those that use EMC storage and those that use Storage Service Providers (SSPs) that use EMC storage - this view is inconsistent with reality. The market for storage technology is bifurcated: On one side are the traditionalists, who seem content to buy storage based on brand name and reputation; on the other side is a growing population of consumers who are interested in pursuing other, potentially less expensive, and more open, options. EMC currently enjoys a reputation for high-quality service and good software in their products that will ensure them continuing placement high in the food chain of the storage vendor world. However, as the economy slows in 2001, companies will scrutinize their storage infrastructures to see how they can do more with smaller budgets and staff. If a storage network can be created that delivers the same performance as an EMC solution, but at a tenth of the cost, there will doubtless be many companies willing to give it a try.
Storage Service Providers (SSPs) will continue to confront infrastructure limitations. The originally stated goal of SSPs, including StorageNetworks, to provide storage as a utility-like service to customers, began a subtle change in 2000 that continues today. Roll-out of the high-speed networks that SSPs counted on to interconnect the customer with an off-site, SSP-maintained storage pool has slowed considerably. Metro Area Networks operating at core carrier speeds do provide services to businesses located in the major NFL cities, but outlying areas - where the bulk of corporate data centers are located - still lack the necessary connectivity and bandwidth to enable SSPs to deliver on their original value proposition. This fact saw most SSP vendors turning their marketing strategy away from remote storage on demand, toward a greater emphasis on co-location services. Most have teamed with large Internet-hosting service vendors to facilitate shared storage within the world of Web hosting. This trend will not change significantly in 2001.
As belt-tightening takes hold in Corporate America, companies will begin to look at storage costs more critically. Management of storage is costly without effective tools. The efficacy of tools is limited by the diversity of storage platforms, as well as by the physical distribution of storage resources throughout the corporate environment. Yankee Group places current management costs at approximately $13 per GB, per year - $10 higher than SNIA. Contrary to vendor marketing, deploying a new storage topology (such as a SAN) to address this issue is only a partial answer. There is no evidence, for example, in those companies deploying SANs today that IT managers are prepared to rip out all non-SAN storage topologies in the process of fielding the new SAN architecture. Hence, the diversity of topologies for storage, the distribution of storage resources, and the lack of effective cross-topology management tools will continue to drive storage costs high and to the right in 2001. What is really needed is more investigation by companies into the origins of data growth itself. Companies will begin looking at what kind of data they are storing and how much of it really needs to be maintained online. They will look at the efficacy of current data replication strategies and will seek ways to share data at the data level, rather than the data copy level.
Evolution, not revolution. Until now, SANs have been touted as a revolutionary storage strategy. Vendors will be recasting their strategies as evolutionary in 2001, following the lead taken long ago by NetConvergence. Smart vendors will follow the approach of NetConvergence, and now Cisco Systems, to describe methods for evolving a storage infrastructure over time and in a non-disruptive way.
A storage disaster in the offering. Given the exponential growth of storage and the lack of effective management of storage in most organizations, storage is a disaster waiting to happen. IBM Business Continuity and Recovery Services reports that disk demand from companies undertaking disaster recovery planning is growing at a rate of only 15 percent, despite the fact that the same companies are growing internal disk storage at a rate upwards of 80 percent per year. This represents a significant gap in the readiness of organizations to recover data used by production applications if and when an outage occurs. Companies have yet to address how they will expeditiously recover many terabytes of storage in multiple topologies in an expedient way in the wake of an unplanned interruption. Recovery from tape is becoming increasingly less viable because of the protracted timeframes involved. In short, we may well see a disaster scenario in 2001, in which corporate servers and networks are restored within a few hours, but recovery of storage requires several days - if it can be accomplished at all.
Like a box of Valentine sweets, storage in 2001 will comprise a mixture of products and solutions that will be chocolate-coated by vendors, as much to enhance as to conceal their real ingredients. The good news is that the products that prove least appetizing can be sampled, and re-deposited, half-eaten, into their wrappers.
However, storage components, unlike Valentine candies, do not come with a legend of contents on the reverse side of the box. IT managers need to be on guard that they do not accidentally nibble on the chocky with the spring-loaded, cheek-piercing bolts.