Unfulfilled promise of SoIP

The acronym SoIP stands for storage over Internet Protocol. But it relates to a 'story of inhibited potential' – a tale of vested interests, customer apathy and half-hearted vendor support.

Yet when systems and software giant IBM trumpeted the launch of its first IP-based storage device – an array of disks called the IP Storage 200i – in early 2001 as an "industry breakthrough", it seemed as if the technology was poised for explosive growth. Ram Jayam, a vice-president of IP-based storage vendor Adaptec, called SoIP a "humungous opportunity" for the sector. Commentators dubbed it the 'next storage revolution'.

But almost 18 months after IBM's proclamation, SoIP remains a revolution in waiting. Now those same commentators are starting to ask if SoIP is going to become yet another hi-tech innovation that worked in the laboratory but flopped in the market.

From an end-user perspective, it seemed like there were cast-iron business arguments for investing in IP-based storage. By 2005, according to IBM, there will be 13 times more data stored than today. How all that data will be stored, according to IBM, will be a "big issue in the near future".

Applying IP to this task seems, on the face of it, to represent the definitive 'no-brainer'. While virtually every enterprise network in the developed world runs over IP, only about 5% of those companies have built a storage area network – the so-called 'sub network' of dedicated storage devices that can be accessed by any computer or application on the network. Until now, the transport protocol for SANs has been fibre channel, a technology whose products are notoriously expensive and difficult to roll out and support. Until SoIP, the notion of building a SAN remained a luxury for most companies.

SoIP promises to change all that. The ubiquity of IP networks means that if companies without a fibre channel network want to deploy SANs, they can do so relatively painlessly. Hardware will be cheaper and technicians can use their existing skills in IP. "The main benefit [of SoIP] boils down to cost," says Nick Tabellion, chief technology officer at Fujitsu Softek, the Japanese storage software supplier.

Such is the promise of SoIP that several major companies, including IBM and Fujitsu, invested tens of millions of euros into the technology. Their aim is to lure customers away from the fibre channel-based market for SANs.

But while SoIP provides companies with a cheaper method of rolling out SANs, very few are buying. IP-based storage devices, say analysts, are attracting little attention from customers, outside of universities or small businesses. So why has the market failed to ignite?

The first problem: IBM and others read the market – and the economy – wrong. For all the hype and promise, most organisations do not require a SAN, at least in the short term, whether it is IP or fibre channel-based. Critics argue that most enterprises will be able to cope adequately with their existing storage systems, such as direct-attached storage and network-attached storage, at least until the end of the current freeze on IT spending.

But there is another more important reason. Few of the world's leading storage vendors, including EMC, Hewlett-Packard (HP), Hitachi Data Systems (HDS) and Sun Microsystems, have followed IBM's lead into IP-based storage.

Indeed, many are dismissive of SoIP. Most vendors maintain that fibre channel delivers better performance, typically offering twice the data throughput speed of its IP-based counterpart. And there seems little evidence of a determined effort on the part of vendors to improve things.

Jim Rothnie, chief technology officer at EMC, the world's biggest storage systems vendor, does not view SoIP as a significant technology. "I don't think it will cause turmoil in the storage industry," he says. "It is just another [storage] transport mechanism." EMC's great rival, HDS, concurs. "When we believe it is a robust technology then we will start to look at implementing that into a physical product," says Trevor Williams, storage network solutions manager for Europe, Middle East and Africa (EMEA) at HDS. The implication: HDS is unwilling to invest heavily in improving SoIP's performance.

Their attitude is not surprising. IP-based storage devices and network components threaten the status quo – and will require suppliers whose products support the fibre channel protocol to write off those investments. This is particularly worrying given that fibre channel-based SANs have still to make significant inroads into the market, and still command high prices and high margins.

Some companies are enjoying comparative success in pushing fibre-channel based SANs towards the high end of the market. It is a trend that London-listed fibre optic network provider Fibernet, for example, can testify to. Fibernet announced in April 2002 that it had provided a fibre channel-based SAN for Swiss bank Credit Suisse First Boston, having already struck similar deals with other financial institutions such as Standard Chartered, Lazard Brothers and JP Morgan Chase. "We have seen this [SAN] market really take off," said marketing director Nigel Pitcher.

While fibre channel SANs continue to penetrate the high end of the market, SoIP products, in contrast, are cheaper and are likely to be focused at the lower end of the market, at least initially.

All of this led storage vendors to back away from revolutionary claims and talk of co-existence and evolution. EMC, for example, believes that fibre channel and IP will co-exist until 2006 at least. Even Cisco Systems, whose strategy is based around expanding the IP world, is careful about talking about one technology replacing the other. "I want to make clear that SoIP technology is not designed to kill fibre channel," says Bernard Zeutzius, a storage networking product manager for Cisco. "This is not a fight of one technology versus another."

Cisco introduced a storage networking router in 2001 that provides the bridging technology between existing fibre channel-based SANs and IP networks. Like IBM, Cisco is happy to talk up SoIP's prospects. But it is only really dipping its toes into the waters of the IP-based storage market. "I would expect around $5 million [in sales] from our storage router for this year," says Zeutzius. For Cisco, a company with revenues of $22.3 billion (€25.1bn) in 2001, those are tiny rewards. "This [router] is only the first step in our story – it is really only connectivity and plumbing," he says. "What is important is to show that IP storage is viable."

Notwithstanding the efforts of IBM – although it too continues to support fibre channel – SoIP's best short-term bet is for a young vendor to emerge with a product sufficiently compelling to shake the storage market out of its complacency. There is a precedent: EMC, then a much smaller company, sold storage arrays at such good value that big disk drives were killed off in less than a decade.

Without something similar occurring, it could take years for SoIP to penetrate the fibre channel-dominated SAN market. Those start-ups whose business models are based around assumptions of IP-based storage taking off quickly do not have time on their side.

Most analysts say that there are simply not enough IP-based storage products available today to drive a bigger market. Equally, the apathy among many users about the role of SANs means that SoIP providers are struggling.

Company   HQ SoIP   Product  
Adaptec California Host bus adaptor
Alacritech California Network interface card
Cisco Systems California Router
Emulex California Host bus adaptor
IBM New York Array device
Maranti Nets California Switch
Nishan Systs California Switch
Pirus Nets M/chusetts Switch
3Ware California Array device, switch

As a result, analysts expect the market for fibre channel hardware components to dwarf its IP-based counterpart for the next few years. According to Gartner Dataquest, revenues from fibre channel-based switches, host bus adaptors, hubs and routers will grow from $1.4 billion (€1.6bn) in 2000 to $8 billion (€9bn) by 2005. In comparison, sales of IP-based components, such as host bus adaptors, converters and gateways, will be worth 'only' $1.2 billion (€1.4bn) by 2005.

Multiple protocols
There is, however, a revolution of some sort rippling through the storage world. SoIP vendors are integrating fibre channel support into their products – and many fibre channel vendors are doing the opposite.

Most smaller storage vendors are now taking a pragmatic line. Instead of gambling solely on SoIP, they are hedging their bets, working on fibre channel-based components as well as IP equipment and striking partnerships to ensure interoperability across multiple platforms and protocols.

Take the venture-capital backed Nishan Systems of the US, regarded as one of the most important members of the IP-based storage start-up community. It claims to have been the first company to ship SoIP products, at the beginning of 2001. For IP-based storage, it provides three switching devices and one software-management tool. But it has also rolled out multi-protocol storage switches that support both fibre channel and IP.

That strategy has been taken up by some of Nishan's IP-based rivals, including Rhapsody Networks and Maranti Networks. They too have developed components capable of supporting either protocol. And Cisco has developed a router capable of connecting an IP network and a fibre channel-based SAN.

At the same time, some fibre channel proponents, such as Qlogic, a Cisco partner, are now developing IP-based products. Fibre channel-based switch vendors, including Brocade, McData and Inrange, could be vulnerable unless they follow Qlogic's example and introduce some IP capabilities. But Brocade is understood to be planning to do just that. Claus Egge, an analyst with IDC, the market research group, says that he was told by executives at Brocade in recent weeks that the company planned to support multiple protocols in future products. Brocade has not confirmed this publicly, however.

Egge believes that these developments underline that SoIP will "complement" rather than compete with fibre channel. He expects there will be demand for products that improve the performance of networks running both protocols.

Supporting both fibre channel and IP is a strategy not just limited to hardware vendors. Companies developing storage management software for fibre channel-based SANs, including DataCore, Veritas Software, StoreAge and FalconStor Software, all have, or plan to, support SoIP in future versions of their products. For example, DataCore's software is designed to enable a network administrator to pool together storage resources from across a network from a single computer interface. And while most customers will use DataCore's software to share storage data over a fibre channel-based SAN, they can also allocate it out to devices with low-end storage requirements over IP.

With many vendors either hedging their bets or not supporting IP-based storage at all, most analysts remain understandably cautious about the prospects for SoIP.

"It could reduce the potential of the fibre channel market. But it is more a reduction in potential [than a serious threat]," according to James Opfer of Gartner Dataquest. IDC's Egge is even more cautious. "If SoIP does not take off in a couple of years, then I don't think it is going to happen."


Poor performance

In 2001, the emergence of a new technology dramatically changed the potential of storage networks using the Internet protocol (IP) The catalyst that finally delivered storage over IP was the Internet small computer system interface (iSCSI) protocol, a storage-networking standard that transports data over IP networks.

The defining function of SoIP is its capability to transport so-called ‘block level' data – the most widely deployed mechanism for handling business applications – over an IP network. Until now, block level storage has been restricted to storage area networks based on the rival fibre channel protocol.

But SoIP delivers poorer performance. According to experts, data stored on a one-gigabit Ethernet IP network would typically be transmitted at between 35Mbit/sec and 40Mbit/sec – about a third of the data transfer speed of a one-gigabit fibre channel network.

But a technology has emerged to address SoIP's problems. New IP offload engines, developed by a number of niche hardware vendors including Adaptec, Alacritech and Emulex, transfer the IP and iSCSI processing into a computer's hardware – thereby improving its performance. Another boost for SoIP may come with the roll out of the first 10-gigabit Ethernet products later in 2002.



Avatar photo

Ben Rossi

Ben was Vitesse Media's editorial director, leading content creation and editorial strategy across all Vitesse products, including its market-leading B2B and consumer magazines, websites, research and...

Related Topics