Fibre Channel To Peacefully Co-Exist With IP Storage.

Once deemed a chimera, storage networking on standard IP networks is now a thriving reality whose presence may mark a turning point in the world of storage connectivity options. IP Storage addresses a company's critical need for extended storage connectivity by enabling it to carry out high performance storage data movement like transfers and replication using pre-existing data infrastructures. Not only does this facilitate management and ensure security, but also enables companies to perform MAN and WAN communications to data storage at a lower cost.

The reception from the connectivity community is a mixed one: given IP is extremely robust in conjunction with Gigabit Ethernet, Ethernet companies can look forward to increased marketability. Though the FC end device market will still have a considerable presence, the FC switch market is impacted by having to share their piece of the pie with larger Ethernet/IP vendors. FC hardware manufacturers have not yet developed economies that would significantly reduce per-port implementation costs. Gigabit Ethernet continues to be cost effective on a cost-per-port basis and shows efficient performance metrics. From a user perspective, network administrators have been waiting impatiently for this development, and storage administrators are relieved that storage over IP implements pre-existing infrastructures.

The two technology approaches are FC/IP, an encapsulation technique also referred to, as tunneling, and iSCSI protocol, a SCSI transport technology that runs on top of TCP/IP with no Fibre Channel legacy in the protocol. It does not attempt to support any higher layer protocol such as SRDF. iETF has proposed iSCSI as a standard to move block data over IP networks.

Tunneling

"Tunneling is the best and worst of both worlds," Simon Fok, CEO of NetConvergence, said. Though some may balk at the idea of being locked into FC technology, benefits include the ability to transport data longer distances and to take advantage of the IP infrastructure.

Tunneling is a non-native approach that links one FC SAN to another. The FC routing protocol (FSPF) lets the user know which FC devices talk to one another. The IP routing protocol sets up the tunnel (FSPF or Fabric Shortest Path First) and the FC tunneling hardware will read FC and store the information inside an IP packet. The packet will then run that from the tunnel entry point to the end point where the information will be unwrapped and read. The visibility of FC information is limited to the end of the tunnel. Rather than acting as integral parts of the IP network, the Fibre Channel devices act as an overlay. One possible consequence of this is that companies will incur overhead from hiring two people to maintain both FC and IP network infrastructures.

According to Mark Knittel, group vice president of worldwide product operations at CNT, FC/IP will allow the fast deployment of IP to interconnect FC SAN. CNT is currently in the process of using tunneling for point to multi-point data transportation.

One of the main drawbacks of tunneling is the lack of speed. FC runs at 100MB/sec, but because the FC information must be packaged prior to entering the tunnel and then disassembled upon emerging from the other end, tunneling is typically being installed at Fast Ethernet (100Mbps) or OC3 (155Mbps). Additional concerns include whether or not tunneling will be able to support the amount of data that needs to be moved during the prescribed backup window.

Regardless, tunneling's quick and easy implementation makes it a potential choice for small businesses with one FC SAN that are not looking to scale to more than two sites. Ultimately, tunneling may prove too cumbersome, and companies will most likely opt for a single resource and a single technology. "I think tunneling will be used in the short term, but over time, I don't know how long it will last," Gary Orenstein, director of marketing at Nishan Systems, said. "Tunneling could become obsolete."

Native IP

Compared to tunneling, Native IP is a more sophisticated technology that appeals to customers who want to integrate native IP end devices, according to Orenstein. The two native IP approaches are iSCSI (new Ethernet end devices that rebuild SCSI serialization from the group up) and iFCP (preserves the standardized SCSI serialization that is part of Fibre Channel and allows users to maintain all Fibre Channel end devices, but link them with IP networks).

The iSCSI protocol is a SCSI transport technology that runs on top of TCP/IP with no Fibre Channel legacy in the protocol. It does not attempt to support any higher layer SAN functionality based on FC. iSCSI protocol is designed to go from Ethernet server to Ethernet storage device, which means that IT managers will be required to add new driver software to Ethernet adapters or replace their storage end device systems. For this reason it seems likely to be a few years before iSCSI will be fully embraced.

iFCP connects FC devices over IP. This is a visible approach that comprises one integrated network and one network routing method using OSPF (Open Shortest Path First) algorithms. On the FC side, the technology implements an IP storage switch that accepts the FC frame data and then maps it to an IP packet. Ultimately the IP packet maintains the address information of FC origin, such that IP network decisions can be made based upon the end device address. The FC information, as part of an IP packet, is transported over any IP network, which could be a shared or separately dedicated IP storage network.

This approach fits the criteria for companies that may not have FC switches, but plan to have multiple sites and a large switch fabric, and the desire to incorporate all available Fibre Channel end devices, such as HBAs, RAID systems, tape libraries, etc., into an IP storage network.

The emergence of Gigabit Ethernet may have a positive impact on the IP storage market. Once the Ethernet community upgrades to 10Gbit and people have faster LANS in place, there will be a performance incentive to upgrade to IP storage.

"I believe that the IP storage market will be extremely successful, even larger than Fibre Channel," Nishan's Orenstein said. Because building a FC tunneling solution is predicated on having a FC network, including switches, FCP over IP and Native IP are potential orders of magnitude in marketability of storage networking. Nishan, citing information from IDC, estimates that the installed base of Fibre Channel switches and hubs is currently at $1 billion. If all DAS were SANs, the value of SANs would be at $9 billion. Morgan Keegan analyst Steve Denegri estimates the market for a device that could translate between Fibre Channel and Ethernet could reach as much as $3 billion annually. Steve Duplessie, a senior analyst with Enterprise Storage Group, puts it at $5 billion today, growing three- or four-fold in five years.

IT managers may well embrace IP storage in the not so distant future. Tunneling could fade out in the wake of IP Storage while Native IP may continue to flourish. And though FC vendors may feel threatened by their Gigabit Ethernet counterpart, they too will reap benefits. Overall, IP Storage appears to be a win-win situation.