It's hard to over-emphasize the importance of disaster recovery in the post-9/11 world. By no coincidence, interest in storage area networks has risen among enterprises looking to network their databases. And as interest grows, storage solutions are evolving from dumb storage array or feature-slim NAS systems to sophisticated solutions where factors such as quality of service, features, performance, availability, network support, and standards adherence influence purchase decisions.
The demand for SAN connectivity and the inherent challenges in DIY systems present an opportunity for service providers to add storage to their corporate service portfolios, says Billy Basu, senior business development manager for optical networks at Nortel Networks.
"One of the remaining headaches for enterprises in terms of operational costs is that they need a dedicated routing network for storage, so they essentially have to run their own router network," says Basu. "That means buying protocol converters and leased lines, which is far more expensive for the enterprise than for the telco offering the service. An additional issue with leased fines is Bandwidth--a T1/E1 connection looks pretty fast for things like Web surfing, but for SAN data traffic it's slow. Imagine copying a 10GB database across a network with a T1 line on either end.
For telcos who want to offer SAN services, metro Ethernet can deliver the access bandwidth that SANs need to be attractive. The problem with Ethernet, however, Basu says, is that it's bursty, which makes it tricky for existing storage protocols like Fiber Channel.
"Fiber Channel is a credits-based system--transmit a frame and a credit is recorded," says Basu. "This means good flow control, but if you're using IP, packets will get dropped every time the traffic bursts, In Fiber Channel, that means an unstable fabric, which means the SAN is effectively down."
This is one reason of several why Fiber Channel has been something of a mixed bag as the de facto storage protocol. It's reliable and scalable, but expensive to install and maintain. It also has a history of interoperability problems.
However many vendors say it's the metro Sonet/SDH network that has to change if carriers are serious about cashing in on SAN demand. Some players suggest upgrading Sonet/SDH with GFP--Generic Framing Procedure, a standard defined under ITU G.7041 originally co-invented by Nortel and Lucent Technologies that gives Sonet/SDH the flexibility to handle both variable and fixed-packet transport. However, with new IP-based alternatives to traditional Fiber Channel coming onto the scene, some carriers may be tempted to skip the interim tech and deploy metro DWDM.
Transparent GFP
GFP is in essence a simple protocol-independent standard frame delineation and encapsulation mechanism for mapping Layer 1 and Layer 2 protocols such as Ethernet and Fiber Channel into Sonet/SDH of optical transport networks. This allows carriers to map packet data into arbitrarily sized TDM pipes--for example, taking bursty Ethernet traffic and mapping it into a fixed bandwidth channel.
In general terms, GFP is touted as a way for carriers to cost-effectively develop metro and wide-area multiservice platforms for rolling out new data services over legacy Sonet/SDH infrastructure without the need for an entire network upgrade.
However, a version of GFP called Transparent GFP (GFP-T) is designed to handle storage protocols like Fiber Channel, ESCON and FICON by acting as a sort of extension cord through the network. Data can be split up and redefined at the far end, but there is no packet processing and no termination of the data, so flow control mechanisms such as Fiber Channel are passed through transparently.
What that means for SANs, says Nortel's Basu, is that GFP gives a carrier's Sonet/SDH network the ability to offer SAN services over Ethernet at bandwidth levels that enterprises want.
"For example, if you have a 10GB database and you need to transmit data from one database to another, the service provider can offer an SLA based on how fast the database is updated, or the speed of the transmission," Basu says, "Just take out the routers and use GFP."
Another key enabler that GFP offers, Basu adds, is that it can expand the reach of SANs over much longer distances.
"You can't do long-distance SANs right now because everyone's doing it with routers and protocol converters," he says. "All those protocol conversions get in the way."
Adoption of GFP is a bit forward-looking, since GFP-compliant products are hard to come by, says Yankee Group senior analyst Pat Matthews.
"GFP is still in its infancy," he observes "For many carriers focusing on their current infrastructure and services is far more important than rolling out new services via GFP. Before implementing GFP, carriers will also have to deal with how to offer the new services."
Matthews also points out that while GFP brings carriers benefits such as the ability to launch new services while avoiding interoperability issues surrounding proprietary bandwidth framing, and mapping techniques with]n the same networks (and in the process allowing them to prolong the life of their current Sonet/SDH infrastructure), it has some drawbacks as well,