Thanks, David.
Most of panelists said iSCSI really "a few years away".
That jives with what the market data is showing.
It now looks like standardized iSCSI smart NICs will come out next year at ONLY 50%-75% of the price of FC HBAs and at ONLY 30%-50% of the performance (1Gbps vs 2Gbps) even assuming no growing pains. This defeats the very business purpose for any mid-range to high-end company to invest in a SAN (a dedicated high performance storage network).
Gigabit Ethernet comprised ONLY 26% of the 3rd Quarter's total networking revenue of $2.5B indicating that the migration from 100Mbps to 1Gbps Ethernet is going to take much longer than the rosier forecasts earlier this year. This effectively undercuts the economies of scale argument that iSCSI promoters have used gratuitously.
10Gbps Ethernet estimates have been seriously ratched down with IDC now expecting 10Gbps to take longer to reach the mainstream. IDC now expects the number of ports to go up from 16,000 ports in 2001 to ONLY 387,000 ports in 2005! This is down from the 4M port by 2005 forecast earlier this year. Not surprisingly, the 10Gbps price per port will only go down from $39,000/port in 2001 to $7,800/port in 2005! Again, even if you assume generously that 10Gbps goes down to $3,900/port in 2005, this still undercuts the economies of scale argument used by the iSCSI promoters because today's FC HBAs are already below $1,000 and director-class ports are already at $3,000/port.
nwfusion.com
Even Cisco is coming around to this view; although in this article, Cree is still being too optimistic about the uptake for iSCSI, which by the way, will likely coincide with the introduction of first-generation Infiniband and Wintel's second-generation 64-bit platform. One doesn't have to be a technical sort of bloke to see that kind of technical logjam created by introducing too many new and unstable technologies at the same time will most likely confuse the customer base and force them to slow down their deployments.
....Cree believes that the uptake of IP storage in big data centers will not take off until 10-Gbit/s Ethernet hits the market. “There are lots of startups developing products here, but it’s going to take two to three years to get into these and then a couple of years after that before it really takes off."
byteandswitch.com
Meanwhile, the FC storage networking supply line appear poised to go from 1Gbps to 2Gbps (by mid-2002) to 10Gbps -- in lockstep with Infiniband and 10Gbps -- and with a 4Gbps option being pushed by Seagate, the dominant enterprise disk drive supplier with 45% of the 2000 market and formidable leads in drive capacity (181GB) and drive speed (15k rpm).
The arguable points are two-fold:
First, the faster deployments of standardized SANs will benefit new technologies like Infiniband and Mckinley the most. Estimates of the market penetration rate of SANs range from 15% (Peripheral Concepts) to 19% (FCIA) so this is clearly a technology poised to move into the mainstream
and stupid vaporware tricks will now only look, well, stupid.
And secondly, if nothing else, this severe IT downturn should make everybody in the IT business more aware to the fact that the dark side of the genuinely accelerated pace of technological advancement today lies in the fundamental limits of the customer base -- low-end to high-end -- to absorb all the fruits of that innovation effectively.
That said, according to this article, there are now more than 300 pre-revenue storage networking start-ups waiting for their IPOs or waiting to be bought out. One doesn't have to go too far back to appreciate how that kind of congested pipeline can create a reality of its own, one where the sky is not necessarily blue.<g>
byteandswitch.com |
