In 1985, the focus of development was System Fault Tolerant (SFT) Netware, a product Novell described as the second "milestone" in the history of LANs. The term "System Fault Tolerant" refers to the ability of the Netware operating system to continue functioning even in the presence of faults or failures in parts of the network. System Fault Tolerance was billed as a milestone achievement because it allowed LANs to accommodate greater numbers of PCs networked in more complicated configurations and still maintain good reliability. And obviously, if LANs were ever to be taken seriously by the business customer, they had to work reliably with a minimum of downtime and provide security against accidental data loss.
The concept of System Fault Tolerance, as applied to Netware LANs, was born in a conversation between Drew Major and Roger White in December 1981. Major said he first appreciated the importance of System Fault Tolerance as he explained the philosophy of Netware in the conversation with White. After subsequent discussions with the other members of Superset, Craig Burton, and others, System Fault Tolerance became a goal of the Netware development team.
The idea was also appealing to Ray Noorda. Before he started looking at Novell in the fall of 1982, Ray had founded a company whose purpose was to develop an inexpensive way to create fault tolerance in minicomputer systems. In a sense, it was the notion of inexpensive, fault tolerant LANs that attracted him to Novell.
In the LAN Times article announcing SFT Netware,* ["System Fault Tolerant Netware, by Michael Durr, page 1, LT, vol. 1, Nos. 3-4, Fall 1985], writer Mike Durr described the market need the new products were designed to meet:
The traditional way to have fault tolerance is to use redundant (duplicate) components. In the mainframe world this may mean installing two mainframes, with the second computer available to take over if the first computer fails.
Minicomputer vendors have come up with another solution that involves building special, fault tolerant computers. Such a machine is equipped with redundant components such as processors, circuits, or power supplies. The machine is highly resistant to system failure. But it is expensive -- as much as $100,000 to $300,000 per mini.
Netware LANs were marketed as a data processing environment that offered inexpensive fault tolerance.
Superset designed SFT Netware as "an extension of the Advanced Netware operating system." Three levels of SFT Netware were to be offered:
Level I: The directory and file allocation table were automatically backed up on two duplicate areas of the network hard disk. (Previously, data was stored in only one area of the disk.) Also, a feature called "Hot Fix" detected bad areas of a deteriorating disk and automatically stored data elsewhere in a good section of the disk.
Level II: Two options were to be available with SFT Netware Level II: disk mirroring and disk duplexing. Both options use duplicate or redundant system components to achieve extra protection -- if the first disk failed, the second would take over with no data loss. In disk mirroring, the second disk is an exact reflection of the first disk. In disk duplexing, the entire disk subsystem (including disk controller, power supply, and cabling) is duplicated.
Level III: In Level III, the entire file server is duplicated, not just components. The file server hardware was interconnected by a high speed bus, and a parity checking feature made sure both servers were sharing the same information.
Another feature of SFT Netware under development in 1985 was the Netware Transaction Tracking System (TTS). This feature, offered with SFT Netware Levels II and III, also helped protect the integrity of the LAN data base in the event of system failure. As explained in LAN Times:
When a database is being updated, if part of the system fails the update may be only partially completed. When this occurs, the database is often corrupted and all of the data in the database is lost.
TTS prevents this kind of data loss by monitoring each transaction (update) as it takes place. If the transaction fails to be fully completed, TTS will automatically roll back to the point prior to the beginning of the transaction. The integrity of the database is thereby maintained. ["System Fault Tolerant Netware, by Michael Durr, page 1, LT, vol. 1, Nos. 3-4, Fall 1985]
SFT Netware with the Netware Transaction Tracking System was indeed a milestone for Novell and the LAN industry. It was a step that had to happen if LANs were ever to be used for the "mission critical" business applications then being handled only by mainframes and minicomputers.