IP storage on the WAN

Save it for the WAN

Fibre Channel may be more efficient for short distances, but over the long haul, nothing is as practical as IP even if it's for functions like disaster recovery. Here, we discuss the possibilities CIOs can explore for SAN-SAN or SAN-NAS connectivity. by Deepali Gupta

The terrorist attack on the WTO twin towers in New York made people think about the importance of backups and data mirroring. A series of natural calamities followed, reinforcing the feeling that Disaster Recovery (DR) and multiple data centres were imperative for the immunity of the business, because as they say, 'the show must go on.'

This is more so in a data-sensitive industry such as BFSI or BPO, where multi-location data-centre-based operations are common. In manufacturing, SANs or storage locations for the factory, branches and data centre can be far apart. Once the storage is in place, the CIO's concern is to ensure that all the storage devices can communicate.

Exit FCP, enter FCIP

In the first phase of SAN evolution, Fibre Channel Protocol (FCP) was the only means for a SAN to talk to another SAN or a NAS. However, pure fibre channel was not feasible for three reasons.

The first was distance limitations; even though Brocade presently promotes native fibre channel that can extend to 500 km, thanks to a newly launched switch family, the standard distance that can be covered using fibre channel ranges between 40 and 60 km.

The second reason is the cost and difficulty in managing and laying fibre over long distances. Not only is the initial cost of the fibre intimidating, but because its tensile strength is low, it is prone to physical damage, particularly in Indian conditions. Thirdly, the central section of the optical channel is never a perfect cylinder. Signal leakage on account of light (signal) emitted through the cable is a problem. While this loss is negligible over short distances, over the long haul, it becomes significant.

A solution that would use copper for long-distance communication between storage devices was required. As most companies already had connectivity infrastructure in place, IP was the logical protocol to adopt for long-distance data transmission. Nevertheless, FCP cannot be abandoned. Organisations had already invested in FCP-compatible storage, and FCP was vendor-specific.

This was good news for vendors, because it meant a brand lock-in for any buyer. As a result, like they had done with SCSI and iSCSI, vendors decided the best way to deal with it was to encapsulate FCP in IP packets and transfer them using the infrastructure for IP. This lead to the advent of Fibre Channel over IP (FCIP) and Internet Fibre Channel Protocol (iFCP). Using WANs for storage data transmission works for enterprises as their IT departments are already familiar with IP and WAN technology. The TCO of an IP-based technology is lower than FC.

FCIP or IFCP

FCIP can only be used to connect Fibre Channel SANs. iFCP maps FC onto IP, but FCIP tunnels data over IP. This makes FCIP a higher bandwidth consumer.

FCIP encapsulates a FC frame within an IP packet quite like IP encapsulates an Ethernet Frame. It is simple and can be implemented in standard switches and routers. On the other hand, iFCP is a gateway-to-gateway approach. Here, the FC frames are mapped into IP packets through the first gateway, and the reverse takes place when the IP packets are received.

In the iFCP market, McData is the only player through its acquisition of Nishan Systems. FCIP-enabled products are available with a number of vendors. See Table 'FCIP/iFCP available in the market.'

Cause for concern

Reliability of WAN for critical data transfer such as disaster recovery will be a cause for concern for most CIOs. In addition, these are bandwidth hogs, which means that most mid-sized enterprises will need to invest in additional bandwidth.

According to Sanjay Kharade, Principal Consultant, Cisco Systems, India and SAARC, "Traditional TCP has a tendency to overreact to packet drops. As a result, reduction in speed or throttling back that occurs is unacceptable to storage traffic." These issues, however, can be dealt with.

The TCP stack can be optimised to carry storage traffic, by adjustment of parameters such as window size, maximum bandwidth, minimum available bandwidth, round trip time and congestion window monitor burst size, Kharade suggests.

One way to ensure WAN reliability is to incorporate a WAN accelerator device. B Chandrashekar, Country Manager, India, Intransa says, "These devices manage WAN band width utilisation to improve performance and different flavours, and provide compression."

Why use FCIP

FCIP has some significant advantages. It is an IETF standards-based protocol and it preserves the FC infrastructure that a company may already have invested in. It is transparent to all existing FC applications and it leverages the widespread reach and economics of IP networks that include E1 and E3 networks. For more information on the various new aspects to storage data transmission that IP has brought in, refer to box 'Opening new doors.'

While some vendors believe the trend of using IP to connect storage devices such as SAN and NAS has already caught on, others like P K Gupta of EMC believe that the technology is "still in infancy, and finding it hard to make a mark in the enterprise set-up." Nevertheless, FCIP and iFCP hold substantial potential for CIOs who want to squeeze the last dollar from their storage investments.

Deepali Gupta can be reached at deepali@networkmagazineindia.com