14.5 Contributing Technologies

The future of storage networking is also being shaped by emergent technologies having only tangential associations with storage. InfiniBand, for example, is first appearing as a server architecture, despite the InfiniBand specification's declared intent to rule the IT world. InfiniBand resolves the limitations of server bus architectures and expedites delivery of application data into server memory. Although it is optimized for high-performance server cluster applications, it has not demonstrated clear end-user value for networking or storage applications. The translation mechanism between InfiniBand and other architectures is defined in the target channel adapter (TCA) specification, and TCAs may provide the link between InfiniBand clusters and Fibre Channel or IP SAN storage assets.

In the Ethernet world, the availability of 10Gbps Ethernet interfaces encourages development of iSCSI storage targets with multiple, high-performance links to serve hundreds of iSCSI initiators simultaneously. Although 10Gbps Fibre Channel may also provide high-performance connectivity, it lacks the ability to seamlessly integrate with mainstream networks and leverage 10Gbps interswitch links now available on Gigabit Ethernet switches. Scalability and simplified deployment, as well as reduced cost due to volume production, are thus giving the advantage to 10Gbps Ethernet.

Higher-capacity and significantly cheaper ATA disks are already exerting pressure on more expensive SCSI and Fibre Channel arrays for midrange applications, and new solid state technologies may displace spinning media altogether at the high end. Enterprise networks that are currently spending several million dollars for high-end arrays may find it difficult to justify the expense for 99.999% up time, when 99.99% can be had for a fraction of the cost. At the same time, the ability to recycle SCSI or ATA drives via cheap iSCSI bridges simplifies the buying decision dramatically for departmental and low-end storage needs.

Object-based storage, distributed file systems, virtual interface (VI) protocols, and remote direct memory access (RDMA) protocols as well as technologies still in deep stealth mode will add to the diversity (and confusion) of the storage networking market. For some vendors, these new technologies introduce chaos and threaten established market share; for others, they provide new opportunities against the incumbents. For customers who can navigate the oncoming waves of technology and can separate vendor marketing from reality, the destination will be, if not exactly the promised land of storage, at least significantly better than the one they left behind.