Network Computing: Bluebird on the Wire
IBM?s ingenious new technology offers network computing to OS/2 Warp Server users. Terence Green explains the benefits
In April this year, OS/2 celebrated its 10th birthday with a new emphasis on network computing which reveals that old dogs can indeed learn new tricks.
IBM?s Bluebird technology expands on the capabilities of OS/2 Warp Server to optimise it for server-centric network computing. The aim is centralising management to reduce computing costs.
Bluebird will eventually be packaged as a set of automated services, but the basic scheme is being rolled out at dozens of IBM customer sites. This is possible because Bluebird is largely based on existing OS/2 Warp Server features.
A Warp Server client can run Dos, Windows, OS/2 and Java applications. The clients can run most operating systems but the OS/2 Warp clients, which are booted from the server, offer the complete central control. A Warp boot image held on the server is used to boot the client workstation, which therefore needs no hard disk ? although if one is present it can be used for cacheing.
In this configuration, all upgrades occur once only on the server, whether for the operating system or the applications. This contrasts with the Net PC, which essentially automates the process of upgrading client workstations by booting them from their local hard disks. At the Net PC launch, an Intel executive described the process as ?gas pumping? operating systems to the Net PCs.
The Bluebird remote boot approach has obvious management benefits and therefore cost efficiencies over the Net PC remote upgrade model. If for any reason, from hardware failure to theft, a particular Bluebird client workstation becomes unusable, it is a simple matter to replace the hardware. No data is lost and no reconfiguration is required, except to update the client?s network address at the server with that of the new network adaptor card.
Warp Server can also function as a remote boot server for Dos and Windows 3.1 clients, so they can be included in a centralised management scheme. Additionally, the latest 32-bit Win 32 applications can be delivered to the Bluebird client workstations using existing Citrix Winframe technology. Win32 applications run on a Windows NT Winframe server and are served up to clients in the same way as for Windows and other clients today.
Microsoft?s licensing of Winframe Multiwin techno-logy, for inclusion in Windows NT 5, won?t affect this capabil-ity since Citrix ICA is one of the two protocols that Microsoft will support.
Nothing in the Bluebird approach prevents full-blown PCs from being used as clients. A percentage of users will always require more power and as Warp Server supports clients running Dos, Windows 3.1, Windows 95, Windows NT, Mac OS and Unix (the last two are extra-cost options), there is full flexibility to integrate existing users in a measured transition to networked computing.
Centralised management and the Java environment, which provides platform independence for clients and servers, form the basis of the network computing model. Java applications are written once and run everywhere; server-centric network computing enables application and operating system up-grades to be applied once at the server and to become available to every client.
Unfortunately, most companies cannot simply uproot existing systems in order to replace them with network computers. Nor are there yet sufficient packaged Java applications to support all the functions that the average business user requires.
Ideally, network computing needs to be phased in, using platforms which run on existing hardware and support existing applications. OS/2 Warp Server and Bluebird techno-logy fits this model very well.
This is no accident: IBM has always adopted a top-down approach. Upstarts Novell, Mic-rosoft and Intel approached computing from the desktop and built upwards from there. That is why they are now having to retro-fit PCs and Windows with systems management tools, which they never considered as important as glitzy interfaces and cool applications.
It could be argued that Microsoft developed the first ?push? technology as ever larger applications and office suites were pushed onto desktops. When the desktop client began to groan under the weight, it was time to buy a PC powered by a faster Intel CPU. When the operating system started to run out of steam it was time to upgrade to a better Windows operating system. It also just so happens that new Windows operating systems often work best with more Intel CPU power and there is usually a new Office suite that you really need in order to exploit the potential of the new system.
Network computing aims to give each user the right amount of computer power, without occupying every desktop in the company with Pentium PCs that spend most of their time waiting for input, and loading them with vast office suites, of which users make use of only about 20 per cent of the features.
This can be done without compromising the user?s ability to do useful work and without forcing all users to make do with a lowest common denominator set of tools. It does, however, demand a centralised approach to systems management to limit the user?s ability to inadvertently reconfigure their system.
PC computing is focused on the desktop user and independence is everything: users are empowered by the information at their fingertips. But this can result in huge support costs. Backing up is an obvious example. With company data spread all over the place, on workstations as well as on file servers, backing up is hard to do. Likewise, the ability to point and click and rearrange your system almost infinitely creates big problems for support teams.
With the space taken up by office suites now approaching 200Mb, even a small configuration issue can take hours to resolve. It is often easier to reinstall the software than to try and resolve the problem. If the user has remembered to back up regularly no data will be lost, but if it is lost, this adds further costs to maintenance.
It was precisely the cost of administering and managing applications over a PC?s lifetime that gave rise to the idea of network computers in the first instance. The basic idea of what is now familiarly called total cost of ownership (TCO) has been embraced by Microsoft and Intel as the underlying principle of the Net PC. Together with zero administration for Windows (ZAW), the Net PC locks down desktops to prevent erroneous reconfiguration and automates the upgrade process for operating systems and applications.
Essentially, the Net PC is designed to support Micro-soft?s revenue model, which depends on a continuous flow of operating system and office suite upgrades. It also fits in with the Intel plan: in order to implement the Net PC/ZAW TCO plan you have to buy a new Net PC.
Initially, Microsoft and Intel were bitterly opposed to the network computer, not least because Oracle?s Larry Ellison taunted them with a sub-$500 box that would replace the Windows PC. The misplaced focus did the network compu-ter no favours, but the basic idea of server-centric computing and centralised management was inevitable given the anarchy of Windows on the desktop. Microsoft and Intel eventually came around too, but hesitantly, explaining that networked computers would amount to 10 per cent of the market at most, primarily as dumb terminal replacements.
A year ago, if you had asked Intel about NCs the company would have told you it was experimenting with various scenarios in the labs but that most users preferred the power of PCs. But then came the first nod to network computing ? the Net PC announcement (requiring new hardware) and the Wired for Management (new hardware) initiative.
In April this year another step was taken: Intel announced benchmark results that showed Java running fastest on Intel CPUs.
And the latest news is that Intel now believes in the Visual Connected computer ? networked computers rather than network computers. The distinction between networked and network is tiny. The important qualifier is the ?Visual?, meaning 3D graphics (faster CPU required), video conferencing (more CPU) and games (more CPU).
There?s a rich irony in all this. Intel has realised the importance of Java and stated so publicly, yet it still insists on ever faster CPUs at the desktop as a priority. The fact is that as Java matures and the Java Beans distributed component model develops, an increasing number of Java applications will run on the server rather than the desktop. Given the speed at which Java has matured ? two years from announcement to a distributed object model adopted by the entire industry ? the Net PC and its requirement for high-powered desktops may have a short lifespan.
Microsoft?s road to TCO takes in the Net PC and ZAW, but there are important differences between the Net PC and a network computer. For a start, the Net PC does not provide complete centralised management because it is designed to pump out upgrades from central servers to Net PC clients which boot locally.
If you have 1,000 Net PC clients, an upgrade entails feeding out 1,000 copies of the upgrade over the network. The Net PC?s inability to remote boot means that version control, scheduling, backup and network bandwidth are all going to be issues.
Microsoft Systems Manage-ment Server and SMS Installer will pump out the software to the Net PCs, and Microsoft ZAW kits will enforce fixed configurations through scripts and system policies. Sealed system cases and the absence of a floppy drive will prevent unauthorised user intervention. This view of networked computing does more to preserve Microsoft?s and Intel?s business models than it does for TCO.
But why does the Net PC work like this? Why can?t Microsoft do with Windows what IBM is doing with Warp Server and Bluebird ? provide a remote boot capability? The problem is Windows NT, which wasn?t designed to support the essential centralised management requirement, the ability to remote boot client workstations. But in fact, that is not strictly accurate.
Windows NT can remote boot Dos and Windows 3.1 clients. Unfortunately it cannot re-mote boot Windows NT Workstation or Windows 95. Since Microsoft?s preferred operating system for the ZAW-managed Net PC is Windows NT, this poses a slight problem, hence the gas-pumping approach for locally booting Net PCs.
Some people say that the Net PC can remote boot, but this is also inaccurate. What the Net PC can do is to bootstrap (using DHCP) a Dos 7 boot image from Windows NT Server. Having booted Dos, the Net PC can then continue to download and install a new operating system, either Windows 95 or Windows NT. This does not amount to a remote boot capability. Essentially, it is an extension of the so-called Windows 95 remote boot to be found in Windows NT, which is more correctly described as a remote setup capability.
If you know how long it takes to setup a new Windows 95 or Windows NT installation, you already have an idea of how long it would take to ?remote boot? your Net PC in the morning: time enough to drink several cups of coffee, read the newspaper and finish the Times crossword.
With Bluebird automation and the existing capabilities of Warp Server, it is possible to take a much bigger bite out of TCO, and all without having to fork out for the new hardware and software that the Net PC/ZAW approach requires.
Unfortunately, IBM isn?t making a great deal of noise about this outside the circle of its major clients. That?s where it is making money out of OS/2, and that?s where it is concentrating its efforts for the time being. The phrase to watch is ?protecting existing investments in OS/2?, and the challenge is to come up with a new name that will help it dump OS/2 connotations.
This is a shame, since the vast majority of new networked computing business opportunities will be provided by small to medium sized businesses. These are the ones which most need the simplicity of centralised management, not having the IT departments of the major corporates.
But Bluebird at least provides an alternative to the Net PC hype by demonstrating a way of networked computing which provides real centralised management and an effective platform for deploying Java applications. Ironically, the perceived deficiency of OS/2 Warp has always been a lack of applications for it, but as Java advances that problem goes away.
Bluebird may not be everyone?s cup of tea but it does address management issues that really reduce the cost of ownership, and it doesn?t require the purchase of a new hybrid Net PC which may turn out to be a short-term investment as Microsoft further refines its Windows strategy over the next couple of years.