Storage area networks (SANs) will inexorably replace traditional host-attached storage, just as local area networks replaced the earlier host-centric networks of dumb terminals in the '80s. The only question is; how long will it take? Or, from the user's point of view, when should I jump on board?
US IT analyst Dataquest is forecasting that SANs, together with Network-Attached Storage, will jump from virtually zero sales in 1998, to 50 per cent of storage market revenues by 2001. IBM is saying that 70 per cent of its medium and large customers will have a SAN by 2002.
What is the attraction? Nick Charles, Overland Data's European sales director, explains why every storage company is jumping on the SAN bandwagon. 'SANs have so many benefits in terms of protection of data, with the modular approach that means that if one storage subsystem goes down, the other modules intelligently cater for it; the greater speed that a SAN allows; and the convenience of having every server being able to access the data. There is also a benefit of around 25 per cent cost saving, because of the greater efficiency of this architecture.'
The one thing stopping everyone rushing out to obtain these benefits today is that standards specifying how SANs should operate are not yet in place. 'The absence of standards has definitely held back the real SAN explosion,' says Nigel Ghent, EMC UK marketing director. Richard Search, Hitachi Data Systems vice president of storage marketing, agrees. 'The necessary standards, especially management standards, are not there.' So does Ron Riffe, Tivoli business development manager, storage systems. 'There is a lot more hype than there is reality in SANs, and customers are having difficulty in installing them if they try to go it alone.'
Who is responsible for producing these standards? The answer is that no single body is involved: there is a large number of different ones. Ron Riffe says: 'Today, there are no fewer than 10 industry organisations involved in creating standards for storage, storage management, SAN management ,and interoperability.'
Some of these, principally ANSI (American National Standards Institute) and IETF (Internet Engineering Task Force), are formal standards-making bodies. These organisations have methodical procedures for ensuring that everyone affected is consulted, and has their say in the process of hammering out a standard, which is necessarily a long-winded process.
To help these official bodies reach an agreed standard, and to speed up the process, a number of industry consortia have been formed. The most important of these is the Storage Networking Industry Association (SNIA), which covers the whole spectrum of SAN standards.
A similar development took place in the early days of local area networking, when Xerox (which developed the Ethernet standard) and DEC, and Intel (the first large industry players to commit to it) put together a consortium that eventually totalled 250 companies. This body played a big part in transforming Ethernet into an official IEEE standard.
But it takes time. Another industry consortium is FibreAlliance, formed by EMC. It is smaller than SNIA, and is more tightly focussed; its specific objective is to push for the development of a SAN management standard.
According to EMC's Nigel Ghent: 'FibreAlliance has submitted an open standard for consideration by the IETF. That was in March , which shows how long these things take.'
While the process of reaching agreement on standards takes place, putting together a SAN requires an individual approach. According to Chris Atkins, Sun storage marketing manager: 'Today you need to build a test SAN with one of all the components, before buying lots of each of them, and rolling it out.'
Sun is not alone in setting up facilities for building ad hoc SANs. According to Tivoli's Ron Riffe: 'Many vendors are putting together testing labs for interoperability in lieu of standards. IBM has some of the largest testing labs of any vendor in the world, in its Global Services operation.'
The standards bodies and industry consortia are working on standards at several different levels. These can be divided roughly into three: standards concerning cabling; interoperability of hardware components, such as hubs and switches; and software, especially storage management and file sharing software.
This tripartite division can be seen as a three step evolution towards 'true' SANs. Sun's Atkins makes the point that 'according to one definition, we have sold 3 petabytes of SAN storage over the past year. That's end-to-end Fibre Channel connectivity from the server through the storage network: on that definition we have installed 3.5 petabytes of SANs. But if you're talking a switched fabric with Brocade switches, the figure is less, and then with heterogeneous devices and servers, it's less again.'
The first step of this evolution has effectively been taken. FC-AL (Fibre Channel-Arbitrated Loop) standards are now in place, and virtually every server and storage vendor now offers Fibre Channel interfaces for their products. According to Peter Coleman, Gadzoox technical manager: 'With Fibre Channel, from peripheral device to infrastructure, there is an agreed, and adhered to, set of standards. So a host bus adapter from any vendor can be connected to any hub or switch, from any other vendor.'
Vincent Franceschini, Hitachi Data Systems European SAN and high availability solution manager, agrees. 'There has been a lot of development from the Fibre Channel community and the storage vendors. Fibre Channel has at last got the potential to address SAN solutions - a lot of Fibre Channel standards issues have disappeared.'
The second step is under way. According to Franceschini: 'The Fibre Channel switch vendors have an initiative in the US that we approve of. We need standardisation of communications levels - that's what OSFI (Open Standards Fabric Initiative) is about. McData, Gadzoox, Brocade, Ancor are working together, so that customers can plan the move to homogeneous solutions.'
At the highest level, software or application standards, there has so far been little progress. According to Steve Richardson, marketing vice president at Overland Data: 'At the application level, and at some of the architectural level, there is not wide agreement among vendors,' adding: .'There are no official standards bodies working on the application layer: that will really be a tough nut.'
Hitachi's Franceschini agrees. 'The real business stopper is the management of SAN solutions. A lot of customers have management packages, such as CA, Tivoli, Veritas, and OpenView. Customers are demanding that SAN management becomes part of that package. The SNIA is trying to develop that with the CIM model.'
CIM (Common Information Model) has the widest support of the emerging storage management standards. According to Franceschini: 'The purpose of CIM is to provide a mechanism or structure to address all different types of information status - network devices, server devices, SCSI structure. It's a globalisation of the management concept.'
How long will it take before all these necessary standards are in place? According to Sun's Atkins: 'It is a two year programme; it will get better and better.' Others are less optimistic. Overland Data's Steve Richardson believes that 'storing data on multiple heterogeneous systems is years away yet.'
In the meantime, users have to rely on vendors to put together SANs for them, knowing that they will not necessarily comply with the standards that do emerge. According to EMC's Nigel Ghent: 'A number of users have got production SANs in operation, despite the fact that a SAN management standard isn't there. But there has been no widespread take-up, mainly because of the lack of an open standard.'
Tivoli's Ron Riffe argues that 'one of the impacts is that the customer has to test themselves, or rely on a system vendor. They need to find a product that hasn't just picked one standard.'
That's true if you want to implement a truly heterogeneous SAN, with storage devices from many vendors connected to a variety of different hosts. But, as Sun's Chris Atkins points out, many users can gain benefits from SANs today, without going so far, particularly if they want to implement host-free backup.
'That's the way people are putting in SANs, to deal with a small part of their overall storage architecture.'