Last week we reported on what Cisco claimed was an "industry first" and a "technical breakthrough" when it simulated a 3,600-mile storage network in its labs using fibre channel over IP.
It demonstrated a valuable point that storage need not be confined to high-speed dedicated cables stretching a maximum of 60-odd miles but, we argued, it would be more impressive if the network was built for real.
Well, director of technical marketing at Nishan Systems, Tom Clark not only agreed but got in touch to rubbish Cisco claims. Over a year-and-a-half ago, on 24 September 2001, he says, Nishan and a group of participants including Dell, Hitachi, IBM, Intel and Qwest ran a physical system stretching 2,500 miles as an effective storage network using both iFCP and iSCSI protocols and achieved the "technical breakthrough" for themselves.
He's not wrong. It was called the Promontory Project and the network ran from Sunnyvale in California to Newark in New Jersey. It managed a peak throughput of 215MBps and sustained throughput of over 200MBps. "We sent real storage data across it," confirmed Tom.
He scoffs at Cisco's claims and the fact that it was a lab situation. "It's okay to do a simulation but it doesn't really replicate a real network where there are issues with connecting to the infrastructure. You have to deal with real latencies and the volume that you get at those distances." Tom says however that the latencies they found were "better than expected" at just 80ms.
So what was the point of his building the longest-ever physical storage network? "To show the upper limit, to hit the high-water mark." With a system that covers the US coast-to-coast, he says, it shows very clearly that distance is not a factor. It can then be scaled down to whatever needs a company faces.
But why would a company want a storage network several thousand miles long? "Companies need resilient, high-availability networks at an extended distance. It could be that their offices are geographically unstable, socially unstable or politically vulnerable."
Covering such huge distances also means that existing data centres can be used by new offices far away, avoiding the expense of having to build a new one. And, Clark says, companies have started using regional offices to back one another up.
There are other advantages of using IP networks as storage systems. Using the IP infrastructure you are not tied into a particular system. Plus, you do not have to set up a dedicated line to be used exclusively for storage. Both these factors make distance storage far cheaper and hence more affordable for a whole new range of companies.
So, unless anyone can put claim to building a physical system covering several thousand miles earlier than September 2001, we believe Nishan should take the title.
However as Tom himself pointed out, the issues of records is something quite separate. "One of our customers, San Diego Supercomputer Center did us one better last fall by running multiple gigabit lines from San Diego to Baltimore at 740MBps on an OC-192 link," he says. That's an impressive 2,290 miles.
But, even though Cisco's simulated distance covered a larger 3,600 miles, we're still unable to grant it the titles it so generously awarded itself. Nishan apparently demonstrated a 25,000-mile simulation at Storage Networking World last autumn. That's all the way around the circumference of the Earth - at which point of course it would have been far easier to just use a six-foot lead to connect the two machines.