More on the phony SAN vs NAS war:
E-business Hands Storage Industry a License to Print Money
by William Fellows 13 Dec 1999 23:19
Data abhors free disk like nature abhors a vacuum; it expands to fill whatever storage space is available. Suppliers of storage equipment and applications that manage it look as if they have a license to print money if they can stay on top of the game. After all commercial storage technology is not rocket science, as Ron Kilpatrick, general manager of IBM Corp's storage systems division told us recently.
A succession of industry research reports suggest the market for data storage products will rival the market for computer systems in value over the next few years on the back of a compound annual growth rate of 60%. That's a tremendous value opportunity for companies which can become leaders in their field. The valuations of companies which will supply equipment for new generations of storage networks including Brocade, Ancor and Juniper may look stretched, but who worries about valuations any longer if a company is in internet and technology?
It is precisely the internet, but specifically e-business and the volume of data it generates that is driving growth in the storage market. Companies are re-centralizing processing and storage requirements to meet the demands of running and maintaining e-businesses. High availability and continuous availability are key: no availability, no company. But as in all emerging and high-growth sectors there is some confusion as the market plays out and vendors battle over standards, technology and approach.
Suppliers of discrete storage hardware and software products believe users will increasingly purchase storage separately from computer systems. A dedicated storage system will depend less on the server CPU which must handle other tasks in addition to routing data. A storage system can be designed to perform more specific functions. It can be extended and managed independently and eliminates many of the I/O bottlenecks imposed by running storage applications on general purpose operating systems supported by traditional servers.
However server vendors which also sell their own server/storage solutions, including including Sun, HP and IBM, argue they can provide more highly integrated (in Sun's case also more modular) and therefore more reliable end-to-end solutions than the point product vendors. Conversely the likes of EMC, Network Appliance, Legato and Veritas are building businesses whose value depends in large part on the likelihood users will shop at different stores for their systems and storage. In the middle are the Dells and Compaqs of the world, which are selling a mix of home-grown and third party storage products which they certify will work together as a single logical storage solution.
As well as choosing to whether to purchase systems and storage from a single supplier or buying separately, users must also decide how their data is to be marshaled. In the world of high-speed, secure, 24x7 commercial data availability, there are really two alternative storage choices emerging in addition to conventional server/storage combinations. Suppliers of storage area networks (SAN) and network attached storage (NAS) equipment and software each claim they are best placed to service the storage requirements of e-business. Both push intelligence out into the network and use a similar approach of separating the data from the application server itself, which enables data to be managed centrally, storage capacity to be upgraded independently from the server, and data fault-resilience to be built into the system if a server goes down. SAN and NAS, which both, are together expected to account for half of the storage market by 2005 with traditional storage networks accounting for the remainder. Market researcher IDC estimates the NAS market will be worth $5.1bn by 2003, up from $540m last year.
The opportunity for ISVs selling applications that manage these new generations of storage networks is also vast. The market caps of two leading suppliers Veritas Software and Legato Systems, are huge. Indeed Legato Systems believes the market for storage management software alone is growing 35% annually and that it will be a $5bn market by 2003 driven by demand for data availability and protection. These, together with data replication it argues are the killer apps of its industry. Legato says 20% of its business - it did $71.7m last quarter - is now driven directly by companies re-centralizing their storage systems from point-product solutions. It thinks NAS and SANs will be ultimately complimentary not directly competitive. NAS solutions are being built out using TCP/IP over gigabit Ethernet networks which makes them inherently less expensive that SANs, which use faster block-level protocols including SCSI over more expensive Fibre Channel infrastructure.
Network Appliance, which supplies NAS storage systems, believes that as data will tend to fill whatever space is available, it makes better sense to build devices optimized to perform specific storage functions rather than have applications compete for gobs of shared space and I/O on general purpose systems. It accepts the distinction between NAS and SANs will blur over time, driven it says by suppliers - including itself - who will develop fabric (network) equipment and devices which can support both kinds of tasks. Its argument is that while blocking mechanisms such as FibreChannel/SCSI are faster than NAS, they tend to lose metadata and inherently support less intelligence, such as metadata, which means users are likely to end up buying dedicated NAS file system or database servers in any case. Moreover mixing Unix, NT and other systems on a SAN is still highly problematic it notes, as SANs deal with raw data, and can't help users share files between Unix and NT servers.
While there is, it believes, no single silver bullet when it comes to storage, it expects Gigabit Ethernet (NAS) will win out over Fibre Channel given the level of investment equipment makers have already committed. Moreover equipment suppliers have already committed to introduce 10Gb Ethernet products, which means, NetApp concludes, that the speed of LAN networks has effectively caught up with direct network attached storage.
The company isn't eschewing Fibre Channel SANs entirely, indeed NetApp CEO Dan Warmenhoven has said that in reality the industry is "not about network NAS versus SANs, it's NAS plus SAN." NetApp is working with switch vendors to allow customers to reconfigure Fibre Channel networks on the fly to use its filer systems. Its plan is to connect its appliances and application servers to SAN back-ends and overlay them with a layer of functions or "glue" such as distributed lock management for maintaining data integrity which SAN vendors themselves are still in the process of integrating piece by piece with their storage products. NetApp claims it and other NAS vendors already have a slew of data management functions built into their appliances and that by breaking them out, with the help of equipment vendors, they will become available to the industry as discrete services.
At the other end of the scale NetApp is working to make data stored on its appliances more meaningful to the servers they feed. The better data is tailored to meet the parameters required by a specific application application before it actually arrives the less CPU cycles the server will requires to process the data. It envisages NAS appliances storing the entire content of a database and delivering from it specifically formed data feeds to server-based database applications. It admits the concept, referred to as narrow sharing, is counter intuitive, but claims the speed of gigabit networks offsets the downside of putting large volumes of data on to a network. "Narrow sharing is coming," it says, and although it has not made any public indication of a plan to offer such a product for email, it's clearly not far from its thoughts. Dedicated database 'machines' have been tried before but have foundered on the crippling effects of trying to send volumes of data over traditional networks. NetApp points to renewed attempts to create new breed of dedicated appliances such as Oracle's network computer and Raw Iron database server which will need servicing by storage resources.
Additionally NetApp points to the server industry's attempt to define a new channel-based, switched fabric architecture input/output mechanism through the InfiniBand initiative (a convergence of earlier competing Next Generation I/O and Future I/O specifications) as essential to driving convergence in the storage market because it will enable the same devices to support Unix, NT and other operating system environments. The architecture bypasses the underlying operating system and connects the CPU and memory directly to an I/O engine. In any case it suggests perhaps three years from now something other than TCP/IP may be running across Gigabit Ethernet networks; if anyone comes up with an asynchronous blocking protocol for it that is.
EMC Corp, the largest dedicated storage supplier is quickly building a franchise for itself premised on the notion that storage is an independent IT resource and that within five years most customers will not be buying storage as a component of their system. EMC has its feet planted firmly in the high-end commercial market where Fibre Channel SANs are expected to predominate but it also believes there will be a significant revenue opportunity for NAS products that is not to be sniffed at. Its executives refer to the NAS versus SAN battle as a "phony war." EMC's view is that NAS devices are particularly useful for certain dedicated functions, such as file serving and sharing, but are little suited to quickly shifting huge amounts of data as required by online transaction processing applications. Moreover it firmly believes just as it does where conventional server networks or SANs are concerned, that storage shouldn't be built into the application server. EMC's expected to supercede its first-generation NAS device with technology from its recent acquisition of Data General Corp.
Not all of the pieces required to build end-to-end SAN and NAS networks are in place yet. The goal of being able to mix and match devices from different vendors on the same network, provide shared data access from multiple operating systems, and manage all of the storage components from a single application framework are some way off. Steven Bulmer, director of technology at $100m IT services and technology supplier Allied Group Inc believes it unlikely that discrete storage entities are unlikely to be replaced by common storage management services for another four or five years. Meantime the industry is engaged in a typical round of jockeying over standards and protocols which is generally speaking the way companies attempt to foist proprietary mechanisms on everyone else. Nevertheless it's difficult to see how SAN and NAS storage equipment and software suppliers can go wrong, given e-business is still in its infancy.
212.134.35.5
|
