In an article in PC Week yesterday, look closely at what was said about I2O:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Network Operating System Is Covering New Ground
Source: PC Week
PC Week via Individual Inc. : Microsoft Corp. took the wraps off the first beta of Windows NT 5.0 last week, and the next-generation platform, packed with more features and greater standards support, reflects the changing face of NOSes.
As servers have become larger and more critical to the corporate world, networking operating system vendors have cranked up their systems' scalability and reliability, and have added many new features.
PC Week Labs recently evaluated Sun Microsystems Inc.'s Solaris 2.6 (PC Week, Sept. 8, Page 95), which comes bundled with PC connectivity services, and rumor has it that forthcoming versions of Novell Inc.'s IntranetWare will include NetWare Connect, for built-in remote access. The addition of clustering services to Microsoft's new high-end Windows NT Enterprise Server is also a harbinger of things to come.
Several PC-based NOS vendors, including SCO, Novell and Microsoft, plan to add support for hot-plug PCI, which allows PCI bus adapters to be added and removed without shutting down the server.
Further standardization across NOSes will occur with efforts such as I2O, which is being adopted by most vendors of PC-based NOSes to allow device-driver interoperability.
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
In yet another article titled, "Internet Forces a United Front in NOS Territory", the following paragraph appeared:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
This push to adopt standards is even starting to go beyond the Internet. Recent efforts to develop the I2O specification are aimed at harmonizing device drivers between operating systems. Hardware vendors have been complaining loudly about the cost involved in creating drivers for numerous operating systems and desperately want to see a standardized model. In the PC NOS arena, I2O support has been pledged for upcoming releases of IntranetWare, Windows NT and even Unix (from vendors such as SCO and Hewlett-Packard Co.).
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Notice something? Neither article mentioned anything about servers, throughput, or off-loading input/output from CPUs when referencing I2O. The point is that the important thing about I2O is not speed-related, nor does the most important thing about I2O have anything to do with servers per se. After all, IBM, Compaq and Intel, among others, have been doing off-loaded I/O for years. The important thing about I2O is that it relieves the industry from the heavy burden of making software and hardware interoperate across networks and all kinds of devices.
It is necessary that I2O, or any other claimant to the throne, must perforce solve throughput problems with I/O, a growing nightmare on servers. I2O does that splendidly, making it an easy sell for servers. But sufficiency requires that it enhance interoperability, and that it become pervasive, ultimately obviating the need for legacy I/O driver development.
I2O solves the interoperability problem, and by becoming an instant success with servers, I2O has everything needed to become a standard. As an emerging standard, I2O will gravitate to all other devices on the network, which in turn reinforces the standard.
Want more proof? How about INTSs recent claim to have an I2O solution? Remember Michael Greene reported last November that, when he asked at the AEA conference, INTS executives put down I2O as being insignificant. Clearly they would like I2O to go away because of the favorable implications for WIND. Consequently, their announcement is the strongest endorsement imaginable for the inevitability of the looming I2O hegemony. INTS would only recognize I2O, its rival's development, if I2O was certain to succeed, and not only for servers, but for embedded network devices.
By no means will WIND have a piece of everything I2O. They will get a cut from everything Intel and Digital do with I2O, and probably from many others undertaking the production of I2O chips, or chips with an integrated I2O capability. Somehow, we will just have to make do with a bit less than 100% market share, probably.
What kind of numbers are we talking about? Big, very big. Add up the following: the total number of servers times 2 to 16 I2O chips each, plus the total number of PCs, plus the total number of NCs, plus the total number of adapter cards, plus the total number of printers, faxes, tape drives, disk storage devices (specifically RAIDs). Want more? Add in all the hubs, routers and switches. Not big enough? Add in the innumerable future Embedded Internet Devices. All these devices are potential beneficiaries of I2O.
The ubiquity of I2O will not occur immediately, but with the broad beachhead represented by virtually all PC servers manufacturers, some important Unix vendors, and RAID and other device manufacturers, it might spread faster than you might expect.
Allen |
