To: cheryl williamson who wrote (11243 ) 10/9/1998 5:01:00 PM From: rudedog Read Replies (5) | Respond to of 74651
cheryl - In Enterprise computing, RAS (Reliability, Availability, Servicability) is a requirement for any piece of equipment or O/S before it can be a contender. NT just isn't there, yet. I'm going to give you a gold star for this post. Maybe 2 gold stars. First, it is obviously from the heart and does not reference any marketing PR. Second, you have hit on what I believe is the critical discussion relating to NT enterprise capability (RAS). Third, the point about the need to defend the low end and move to the high end all with a single product is the second most important question facing MSFT both from a development and architectural point of view. Jack of all trades is master of none. Current NT Server can achieve decent reliability (99.9) but only in a closely controlled environment. First start with good hardware which can survive failure of key components (disk, NIC, fan or power supply) without telling the OS about it. Compaq, IBM, HP and DG all have systems that fit the bill. Then, only run a system which requires little administrative action - don't change network protocols, don't change disk configuration, don't add or subtract anything which changes the registry. There are many server sites which can fit into this model, but not too many interesting ones. In a general mix of application and 'file & print' servers, most testing I have seen shows reliability of between 95% and 98.5% available. This implies unanticipated outages of hours per week, simply unacceptable for any mission critical tasks. Microsoft is certainly aware of this and I believe they will be able to routinely achieve 99.9 reliability for first tier configurations but not until mid-2000 at the earliest. The project is late, indicating that it may have gone out of control. My guess is that MSFT will come out with a scaled-down version next year 100% accurate IMO. I believe it is accurate to say that MSFT was not prepared to deal with an effort involving nearly 2000 developers and 50 million lines of code, and that some serious realignment of priorities, methods, and development goals has been required to avoid meltdown. The current push behind Linux may give them some added competition, especially if lots of personal-productivity applications get ported or are written for it. The quest to get a clean, small, easy to use consumer OS out of the NT kernel will be a great challenge to MS. They have not only Linux to fend off, but they also have to replace Win9X as soon as possible or they will have osborned themselves in the commercial desktop space. The pressure will be on to jam features into the core OS to save time, the same kind of pressure that created the sloppy Win98 product. It will take a lot of strength on the part of the NT team to keep the feature creep at bay while still driving schedule. I believe that MSFT will need to regroup and develop a different internal culture and development methodology to address the RAS issues, and that they will probably need to develop at least 2 and maybe 3 related but separate NT products (or products intersecting in only a fraction of the codeline) to cover the range of the NT market. I believe that they will successfully do this, and that they will continue their growth into the enterprise space. Their defense of the low end is a little less certain but less important if they can hit the high end Unix space within 2 or 3 years. Thanks again for an interesting post, I must confess I was surprised but I like the new style a lot more than the old, please keep it up.