Sören Janstål

Data Research DPU ab Data Research DPU for Evaluation of Information Technology

Sören Janstål

The advent of the Personal Computer, and more specifically the advent of the Graphical User Interface (GUI), marked a transformation in the usability expected of computer systems. Originally intended as an interface to ‘personal’ applications, the relative simplicity of the GUI encouraged the deployment of PCs for use with shared applications – and gave birth to the Client/Server phenomenon. While a number of different architectures were tried, the technology available caused most implementations to be of the ‘fat-client/thin-server’ variety. This name derives from the client device (the desktop PC) having to absorb the majority of the processing. Typically all applications and business logic execute on the client hardware with the server simply managing the data content.

After the distinct lack of user-friendliness of earlier computer applications, Client/Server was seen to be such an advance that whole corporate IT strategies were built around it before the technology was properly mature. Only after a critical mass of systems had been deployed did the true cost of the fat-client/thin-server model become apparent. The last few years have seen the ‘Total Cost of Ownership’ (TCO) issue become a major controversy, with almost as many solutions proposed as there are IT vendors. Various estimates of TCO using different assumptions and based on different system profiles have given values between $6,000 and $15,000+ per user per year.

Citrix WinFrame is probably the most widely accepted and widely deployed solution to the TCO problem. It works by providing a multi-user environment on a shared server where user sessions execute. The only processing which occurs on the client hardware is the minimal amount necessary to manage the user interaction – the client simply operates as an input/output device in much the same way as ‘dumb terminals’ operate in a legacy Transaction Processing system. No changes are required to the application in order to achieve this, yet the interface that the user sees is indistinguishable from that of the GUI-based fat-client.

The most obvious cost saving is that lower specification (and hence cheaper) client hardware can be deployed. While virtually any client device may be used, low cost specialist devices such as the Wyse Winterm are available and have proved to be an effective option due to the ease of maintenance, greater reliability and reduced opportunity for ‘user-induced errors’. Yet the experience of users is that other savings in terms of reduced management costs and greater user flexibility have still greater value to the organisation. These savings are due to moving the complexity of applications away from the users’ workstations and back under the control of IT departments. We return to this in more detail later in the report.

Wherever there are potential savings there are also potential costs. In this case, the process load is moved from a large number of clients to a central server (or group of servers). Experience shows that this is a good trade-off, both in terms of the cost of the total processing power and in other associated costs. A more subtle objection is that many users like to be regarded as ‘power users’ who need a state-of-the-art PC. In this respect, the PC has usurped the company car as the prime status symbol.  This is an attitude problem which has little to do with the cost-effective use of technology. Other rumoured costs – such as the need for radical upgrades to the network capacity – seem to be largely unfounded.

The Citrix model has now been endorsed by Microsoft through the recent launch of the Microsoft Windows NT Server 4.0 Terminal Server Edition. This product delivers essentially the same multi-user functionality as Citrix WinFrame, and was developed with the cooperation of Citrix. But any organisation intending to deploy more than a trivial number of users, or to provide client devices other than Microsoft Windows PCs (or Windows CE) will certainly need to consider the use of Citrix MetaFrame which provides additional functionality and performance in these areas.

This paper explores the Citrix Thin-Client Server Computing architecture created by Citrix in greater depth and draws on the experiences of BT Syncordia in implementing both Citrix WinFrame software and Wyse Winterm clients. As detailed later, these two complementary technologies have provided significant benefits through direct cost savings and enhanced productivity.

The Pendulum Swings

In the IT industry we always seem to take technologies to extremes. It has been the same for every generation of technology. When Transaction Monitors made online TP applications a possibility we went overboard and started rewriting all our old batch applications to provide an online interface. In many cases, this made little sense as the business process behind the tasks had not altered – but we rewrote them anyway.

At least we started to rewrite them, because just as the pendulum reached full swing towards TP systems, personal computers gave rise to the Graphical User Interface. Users asked why a system which cost hundreds of thousands of pounds to deploy had a user interface entirely inferior to that available on a PC with a total price tag – including software - of less than two thousand pounds. There was no satisfactory answer to the question except that the technology had not been commercially available when the systems were being designed. So we started to see small-scale deployment of applications shared between a few PCs connected via a LAN.

In comparison with what had gone before, users loved the more attractive interface which provided the capability to navigate through systems with a mouse and the greater consistency in the interface between applications from different sources. For the first time most staff were given direct access to personal applications such as word processing and spreadsheets in addition to the applications which directly supported their job functions.

Sensing a new round of purchasing starting, software vendors (encouraged by PC manufacturers) built new versions of their application packages in the Client/Server model, and new development tools emerged to allow user sites to build their own systems cost-effectively. The inertia to deploy Client/Server applications was enormous.

Then the problems started. Small scale deployments seemed to work well enough, but extending these to enterprise-wide applications presented difficulties.

Performance was disappointing – sometimes unacceptable. Worse still, the reliability and availability suffered. And a lot of money seemed to be falling into a black hole. Analysts started to look at where the money was going, and ‘Total Cost of Ownership’ (TCO) became the phrase of the year.

Hardware	Software
Upgrade	Networking
Standardisation	Recovery
Availability	Software Management
Backup	Archiving
Security	Performance
Training	User Support
Software Development	Data Duplication
Process Duplication	Administration

Table 1: Some Costs Associated With “Fat Client/Thin Server”

In fact TCO is almost impossible to define adequately. At Bloor Research we have identified at least eighteen different ways in which Client/Server absorbs money – as shown in the table. It is likely that experienced systems managers will be able to add their own items to the list. Various attempts to quantify these costs have resulted in estimations varying between $6,000 and $15,000 per user per year.

Now the pendulum has started yet another swing. Encouraged by the wide-scale demand for Internet and Intranet applications, and because of the increasing awareness of the cost of the fat-client, the industry trend is accelerating towards thin-client computing. But just as Online TP was not a complete answer to the shortcomings of batch processing, and Client/Server was not a complete replacement for Online TP, so it is highly unlikely that thin-client will provide a complete replacement for the activities that are well handled by the fat-client model. At this point in the cycle we have the opportunity to stop the pendulum from swinging too far towards the opposite extreme. We have the technology to support the sensible choice of client platform depending on the work activities of individual users.

Reducing The Cost of Client/Server

There have been a number of initiatives aimed at bringing the TCO under control. While there are several variants they fall into three significant categories – centralised management of PCs, component-based computing, and thin client computing.

Centralised Management of PCs

Management software to enable the centralised administration of PCs is available both as standalone products and as functions within high-end systems management suites. Functionality varies, but normally includes the ability to remotely monitor the software configuration, perform remote virus checking, and the ability to push upgrades down to unattended PCs to help ease the management of version consistency.

An extreme solution is to combine the remote management capability with the deployment of Net PCs. A Net PC has no diskette or CD drive and is either a sealed-box unit or uses a motherboard with no expansion slots to prevent user tampering. The Net PC has not been widely adopted.

Centralised management can certainly reduce ongoing costs at some initial expense. But the same tasks still exist and must be implemented - all that is really happening is that a degree of automation is taking place.

Component-based Computing

The concept of component-based computing is that applications should consist of small components stored and managed by a server and delivered to a client as needed to satisfy a particular task. The business logic is intended to execute on a server platform, leaving the client to execute just the logic required for display and input (along with basic input validation, etc.). The components are generally in the form of Java Beans, but an alternative model uses Microsoft COM ActiveX controls. While there has been much publicity (both positive and negative) about the Network Computer, the actual client devices employed are less important than the architecture. Some early implementations of component-based computing missed the point that this is really a server based architecture and attempted to deliver large monolithic applications on demand to thin-client devices. This lead to many of the exaggerated claims about the impact of network computing on bandwidth requirements.

Implemented correctly, component-based computing provides one implementation of a thin-client model. Poor implementation re-introduces many of the costs associated with fat-client processing.

Bloor Research believes that component-based computing will become a long lasting and vital architecture – but it does require significant investments to be made in new applications or in major enhancements to existing applications. It will therefore take some years before there are many significant systems deployed in this way.

Thin-Client/Server Computing

The purpose of Thin-Client/Server Computing is to provide an environment where applications can execute on a shared server, similar to the Mainframe concept with the ability to run Windows applications. This is the subject of the next section – the Citrix environment.

There are two central elements addressed by Citrix:

• the requirement to move applications off distributed PCs on to a shared server
• the need to continue to use Windows-based applications on which businesses depend and with which users are familiar.

These two capabilities provide the cost reduction potential of thin-client computing, together with a reduction of the risks associated with any large-scale change in technology. Cost reduction with low risk is a killer combination, and this is precisely what Citrix have achieved. The notion of running ‘personal’ Windows applications in a shared environment is a very powerful idea – hence the term ‘Windows By Other Means’ used in the title of this paper.

Citrix WinFrame has two main software components – a server-based component which provides the execution environment for Windows applications, and a small client-based component which manages the interface across the network.

The server software provides a multi-user environment capable of running Windows 32-bit applications based on Microsoft NT 3.5.1 source code technology. Later releases of NT require Microsoft’s Terminal Server product together with Citrix MetaFrame as described later. The only impact from the users’ perspective is that with Citrix WinFrame the client interface adopts the NT 3.5.1 look and feel – which is essentially the same as that of Windows 3.1.1. This doesn’t impact the functionality of the applications, but users who insist on a Windows 95  style of interface will need to implement Microsoft Terminal Server and Citrix MetaFrame.

The server environment allows multiple users to log on and run applications in separate, protected sessions on a single server. Application errors occurring within one user session will not have any impact on other concurrent users. In the worst situation it will be necessary to stop and restart a user session – a much faster activity than re-booting a PC in the equivalent fat-client situation. The server environment uses Citrix ICA protocol (Independent Computing Architecture) to separate the application logic from the user interface at the lowest possible level of mouse clicks and keyboard strokes. Administration features are provided to manage the user sessions.

Client-side software is available for a variety of devices including PCs running Windows, DOS or Mac OS, Windows-based Terminals (WBTs) such as the Wyse Winterm described later, and any Java-enabled client device (so also including network computers).

It is important to differentiate between ‘network computing’ as an application architecture and ‘network computers’ as thin-client devices that conform to a particular industry specification. This confusion in terms has lead to a great deal of misunderstanding – and in particular has given rise to the notion that the entire current investment in PC client devices and applications must be scrapped in order to participate in the new computing architecture. This is emphatically not the case. Network computing simply refers to a multi-tier application architecture where clients (the input/output devices) are separated from application servers (the business logic) and data servers (databases and files) across a network. The network may be local (a LAN), wide (a WAN) or ‘hyperspace’ (the Internet).

Because the range of client devices include conventional PCs, applications can be moved piecemeal to exploit Citrix capabilities. A PC can still run ‘personal’ applications entirely from the local hard drive, and can still run fat-client/thin-server applications with the business logic executing on the PC – but the organisation can chose which applications would be most cost-effective migrated to a Citrix thin-client architecture. This is particularly valuable when there is a requirement to add a new application to a PC which has no further disk capacity. The small Citrix client environment takes little disk space yet opens up access to any number of new server-based applications.

Because WinFrame will happily coexist with fat-client processing it is likely that most organisations could benefit from deploying it in a selective manner. The only Windows-based applications which are not suitable for execution in a WinFrame environment are those which make explicit use of NT4.0 (or later) features. These are accommodated by Citrix MetaFrame as described later.

Given that any PC is capable of running the Citrix client-side software, it might be assumed that most sites will elect to retain their existing PC investment until old age and maintenance costs make it cost-effective to replace them. Many users will continue to work in this mode, since it gives the ability to retain applications with a heavy processing requirement on the PC to avoid the impact of these on other users. But for users who need to access the more usual data-intensive applications, or for those who need access to ‘back office’ types of products (such as word processing or spreadsheet applications), experience has shown that there are advantages in replacing PCs with more appropriate devices.

In most circumstances a ‘Windows-based Terminal’ – or WBT provides the best solution.

The term WBT covers a range of client devices which provide no user-accessible processing capability and no user-accessible storage. They feature an embedded operating system which supports a communication protocol to transport keystrokes, mouse clicks and screen updates between the client and the server. One of the most popular choices of WBT is the range of Winterm thin-clients from Wyse. The range consists of a spectrum of devices from simple, low cost terminals to multimedia-capable clients or wireless-connected mobile units. The units can take on different ‘personalities’ by loading the appropriate emulator, so the same device can behave as a WBT, UNIX console, VT100, 3270 or other terminal type. Wyse can provide software to centrally manage the emulation mode, and to permit concurrent use of graphical and character-based applications.

But what is the sense in making an additional expenditure to replace functioning PCs with new WBTs?

Reasons will depend on circumstances, the common factor being the greater reliability of the simpler devices. Increased reliability is not just because of the elimination of moving parts. Many support calls for PCs - whether they are used as conventional personal computing devices or as thin-clients – are due to user-initiated problems. In short, users may alter the setup, load unauthorised software, introduce a virus (via diskette, CD or Internet), and even attempt hardware changes themselves. While systems management software can provide a high degree of protection, this too has its costs both in terms of the software licensing and in terms of the effort needed to establish and implement the management policies. Replacing the physical devices is a simpler, and usually more cost-effective, solution.

A further reason to consider replacing existing client devices is to consolidate several different terminals into a single unit. Through years of evolution of IT systems many users have acquired a desk holding two, three or more different terminals to give access to systems – mainframe, VAX, UNIX, PC LAN. Moving to a thin-client architecture provides a convenient opportunity to simplify the entire terminal environment to a single specification across the entire user base. Support calls will be reduced in number and simplified in nature. Because the user profile and data is no longer a property of the client but is stored on the server, failure of a client device can be dealt with immediately by plugging in a replacement unit and fixing the problem while the user carries on with the job in hand.

It should be apparent from the overall architecture that client hardware deployed to meet immediate business needs will almost certainly be appropriate for any future business applications and any future upgrade to the Citrix software. For example, moving to Microsoft Terminal Server and Citrix MetaFrame will have no impact on any of the client devices. While this fact might be taken for granted, it should be contrasted with the extensive review of end user processing requirements which would normally accompany a decision to roll out a new version of operating systems or applications across a large number of users. Only a fundamental change in the type of user interface to business applications is likely to cause Winterm devices to become obsolete. Therefore the long term cost equation will swing still further towards this type of client hardware.

Background

BT Syncordia provides total systems integration services to it’s customers, across the technology spectrum.

It’s skills in Network Centric Environments from Voice and Data Networking mean it is excellently placed to offer customers solutions based on thin-client technology. BT’s appreciation of the vital disciplines necessary for any business-critical deployment, coupled with it’s experience exploiting it’s own and customers infrastructures, mean that it has a unique mix of attributes in this market.

BT Syncordia uses BT’s corporate experience and innovative custom-developed solutions to the benefit of it’s customers, by building a vital understanding of the most effective method of deployment and support internally, as part of it’s service development.

This case study relates to an organisational group of 1200 users across 4 sites in 3 key UK locations – Eaton Court and Network House, both in the Hemel Hempstead area, plus London and Manchester. It supports 12 years of legacy Information Systems, many of which are critical to the operation of the company and it’s customers.

Having already spent £3M upgrading a wide range of low specification ‘legacy’ PCs, in September 1996 BT Syncordia found itself needing to upgrade or replace more than 800 PCs based on 25MHz and 33MHz Intel 486s and running under Windows 3.1. The estimated hardware cost for replacement was between £5-6 Million. In order to better support new and upgraded applications, including improved versions of Microsoft Office, BT also faced an upgrade to Windows 95 for 570 of the desktops – which would carry deployment and development costs of around £15-20 Million.

BT Syncordia identified three alternative approaches to the upgrade:

1.    to replace the PCs with new Pentium-based hardware

Replacement Costs for Desktops		Upgrade Costs for Desktops
570 purchase @ £1250 each    = £712,500		400 purchase @ £110 each     = £44,000
Deployment and Development    = £2,493,750		170 Winterms @ £500 each     = £85,000
		Deployment     = £24,250
Total     = £3,206,250		Total     = £153,250
Deployment Savings = £3,053,000 (Equivalent to £5,356 Per User)

Table 2: Difference of Deployment Costs

2.    to upgrade the PCs with additional main memory, processor overdrives and new cards for video and networking

3.    to replace the PCs with thin-client devices using Citrix software.

Following a successful pilot of the Citrix software (carried out using the pre-upgrade hardware), the solution adopted was a combination of 2 and 3 above with Citrix software implemented throughout. The actual cost comparison for the hardware only is shown in the frame.

Because of the availability of volume purchase arrangements, the PC upgrades were made at a unit cost of £110 per PC. Upgraded PCs also had their diskette drives locked down to prevent unauthorised software being installed. Despite the availability of upgrades many of the PCs were old units, and growing maintenance problems made it more cost-effective to replace 170 of them with new devices, for which BT Syncordia selected Wyse Winterms.

The initial primary aim of the project was to reduce the immediate hardware expenditure, and this was certainly achieved as the figures show. The immediate savings realised in hardware and deployment costs were over £3Million, but BT Syncordia’s own estimation is that subsequent ongoing savings should be substantially higher – in the order of £15-20 Million. On top of these savings, further hardware expenditure has been avoided as some of the PCs successfully upgraded this time would not be capable of upgrading in a year or two due to further deterioration in reliability and escalating support costs.

The Deployment

The physical environment for the deployment at BT Syncordia is represented in Figure 1. There are four sites (fourteen buildings) linked by Fibre Optic cable. Each site has two or more multi-CPU WinFrame server supporting a total of over 400 Wyse Winterm thin-clients.

WinFrame software runs on all the servers in this network and supports all the applications. In addition Windows NT supports the print servers and NetWare is used as the server for email and certain file systems.

The deployment process itself was remarkably problem-free. The Winterm hardware has been proved to be extremely reliable – at the time of writing no hardware failure has occurred on any of the Winterm clients. The only significant hurdle which had to be overcome was not technical; it was user reluctance born out of the expectation of decreased performance and loss of autonomy. Version management was  one activity which improved immediately. Prior to WinFrame deployment there was a mixed bag of software versions across the user base but when users gained access to the system via WinFrame, version management could be applied once with immediate effective for all users.

Experiences

Aside from the quantifiable benefits of reduced costs directly related to hardware acquisition and maintenance, many other benefits were found.

One of the first bonuses was the much improved capability to respond to change. In a conventional Windows environment changes to application software deployed on client devices needs to be carefully planned and scheduled into a quiet period or overnight downtime. Using WinFrame only one copy of an application needs to be deployed in order for every user of the server to automatically use the correct new version. This feature enabled BT Syncordia staff to successfully implement three last minute changes to a 96 user application on the same day it went live. The same features reduced the timescale and staff effort needed to roll out a new version of Microsoft Office across the entire organisation.

Other benefits stemmed from the opportunity given to the IT department to “take back the desktop”. With this application architecture ‘end user anarchy’ simply does not happen. Regaining direct control of the desktop lead to one of the unexpected bonuses – a reduction in the licensing cost of desktop software. Previously more licenses had been acquired than were actually used because it was difficult or impossible to discover which users actively used any particular product. Because all software is managed centrally it becomes a simple task to find which users really need to be licensed.

But probably the most significant cost savings experienced have been due to the much higher reliability of the WinFrame/Winterm combination. BT help desk gathered data over a thirteen month period. The figures showed that as the proportion of WinFrame users increased throughout the roll-out phase, the number of support calls reduced and the proportion of problems fixed by 1^st-line support increased. Very few issues needed to be escalated to 2^nd-level support.

This can be seen in the graph in Figure 2 which shows the number of faults, WinFrame users and non-WinFrame users for each month.

The graph shows that while the number of users was stable, the number of faults decreased as the number of WinFrame users increased.

The superior performance of WinFrame is also shown by comparing the percentages of 1st Line and  Second Line closures for  WinFrame and non-WinFrame as shown in the table:

	WinFrame	non-WinFrame
Second Line	7.65%	21.65%
First Line	92.35%	78.35%

These figures are shown as a graph in Figure 3.

The effect of the introduction of WinFrame was lower direct support costs, lower hidden costs caused by users unable to carry out their tasks, and greater time available to technical staff for carrying out more productive work. This knock-on effect lead to a real improvement in the quality of service to the end users. Initial reluctance on the part of end users to accept the new system disappeared when it was discovered that, far from obstructing them in their work, the new environment made the users’ roles simpler and more productive.

When usability issues did occur (often more to do with the application than with the environment), the shadowing facility provided by Citrix helped provide a fast resolution. Using shadowing a support technician can – with the user’s consent – view exactly the same display as the user, in real-time and from a remote location. If needed the technician can take over running the user session to demonstrate the resolution of a problem. This represents a considerable time saving compared with the need to get a support person on site with the user especially when the problem is trivial. BT Syncordia have provided a cost estimate for an example where shadowing was used to avoid the necessity of sending a support engineer to a customer’s site to fix an urgent usability problem. This single real event saved approximately £3,000 including overtime, travel and disruption to normal work.

The Winterm clients had a double impact on the overall reliability. Experience to-date shows that they are likely to have a mean time between failure of 3-4 times that of the more complex PCs they replaced. Due to their sealed-box nature, they remove the temptation for users to load their own unauthorised software, alter their own profile or attempt to find their own resolution when problems arise. In addition, the Winterms are expected to last for at least 4-5 years in contrast to desktop PCs with a useful working life between 2 and 3 years, and laptop PCs with a useful life of as little as 18 months. Bloor Research believes the actual life of Winterms will exceed this estimation – the unknown factor is not the durability of the hardware but potential future changes to end user processing requirements. In all probability the balance will move further toward server-based processing, leading to greater longevity of this deployment.

Problems which had been anticipated – such as the unknown impact on the network capacity required – turned out not to be problems after all. BT Syncordia actually report a slight reduction in network traffic compared with the fat-client architecture being replaced. The Citrix ICA protocol provides a high degree of compression which means that staff can use the same systems via remote dial-up, and this has become the de-facto standard for remote access.

As mentioned previously, users had been concerned initially about the lack of autonomy and lack of flexibility which would result from the absence of processing power at the desktop. Again, the exact opposite was discovered. In a fat-client scenario the user is entirely dependent on applications – and often data – being resident on the PC. This means that the PC becomes essential to the work environment and has to go where the user goes. If the PC is unavailable for any length of time it would require a replacement device to be configured with the same software and data as the previous PC – provided, of course, that all the required backups were available. This can be a hugely disruptive process and does not fit well with the modern concept of a flexible, mobile work force. Because the WinFrame/Winterm environment does not have any information stored at the client device, using any other client is simply a case of logging on to identify the user. The user’s profile is then immediately available. The flexibility of this ‘hot-desking’ becomes a major productivity enhancement – both in flexible working arrangements and in the event that the client device becomes unserviceable.

Of course there are some work profiles which do need a Personal Computer in its true sense. In the case of BT Syncordia they found that users of CAD applications, and a further set of users who genuinely needed frequent access to CD-ROMs and diskettes were better served by a PC. But even these users can benefit from the Citrix environment, executing shared applications via WinFrame while retaining the PC to run the more demanding applications.

BT Syncordia’s Analysis

BT Syncordia’s own analysis of direct cost savings, as detailed above, is equivalent to more than £5000 per user. But the direct savings are only a part of the story.

User Opinions

Importantly, the reaction of the users involved in the deployment is completely supportive of the technology. The only negative opinion expressed has been the lack of multimedia capabilities. (Multimedia is actually available via WinFrame, and some Winterm models also support its use, but this is an instance where bandwidth requirements would have to be investigated). Users are neutral about the performance issues, experiencing equivalent overall performance to the previous fat-client processing. Particular features highlighted by users as improvements are:

• desk independence (hot-desking)
• central control of backups
• central control of virus security
• reduced desk space requirements.

Business Benefits

In business terms the project has been entirely successful. There was concern that non-Windows based applications may not be supported through the WinFrame architecture, but in fact WinFrame supports a pass-through and the business is successfully running a vital NetWare 3.12 application through both Winterm and PC clients.

The project has highlighted the need to maintain mainframe-quality management of the server environments, since failure of a server will impact many users. It was found, however, that overall availability actually increased. From a business perspective, the benefits identified by BT Syncordia  are:

• direct cash savings of over £5,000 per user
• improved control over software licences
• improved management of application deployment – both in terms of assigning specific users to specific applications, and rapid and controlled deployment of new versions of software
  • more predictable bandwidth requirements.
  • increased availability of systems and less lost time for users
  • improved detection and prevention of virus infections
  • improved remote access to systems
  • enhanced user support and lower help desk cost through ‘shadowing’
  • reduced exposure to theft of PCs (lower risk of theft and reduced impact if theft occurs)
  • prevention of loss of business data by centrally managing backups.

The organisation is continuing to assess new products such as Microsoft NT 4.0 Terminal Server Edition and Citrix MetaFrame, and plans to deploy these where appropriate. Trials of Beta versions of the products have already been carried out with deployment to more than 100 users.

Bloor Research Analysis

The direct savings quantified by BT Syncordia are impressive in their own right, and such savings would inevitably be a key feature of any business plan to implement a thin-client computing architecture. If a cost-justification can be made on these tangible cost savings, then so much the better. But the more important benefits of the architecture are difficult to quantify. Some will remain intangible – such as the improvement in user attitudes due to the improved reliability and usability of systems.

Other intangibles are in the area of risk-avoidance. Specifically, the architecture enforces centralised management of the client environment. It simplifies the task of protecting against virus infections and other problems caused by unauthorised software. It is the belief of Bloor Research that both active ‘cyber-attacks’ or passive viral infections will become a significant threat to organisations of all types, as the means of exposure to such threats will increase along with the connectivity of systems. A thin-client computing model reduces this threat by making it easier to manage.

Other potential savings which are difficult to quantify include reductions in consequential losses due to system unavailability. This will largely be dependent on the type of organisation, but as direct business-to-business electronic commerce increases, the chances of business opportunities being lost through system down-time inevitably will increase. A properly implemented and managed thin-client environment will always exhibit superior availability to an equivalent fat-client model. High availability of systems will become one of the factors which will make an organisation become a business partner of choice. In the case of BT Syncordia, whose business is to provide high-technology solutions to others, it’s own systems become an essential shop window to it’s prospective customers, raising the profile still further.

There are further indirect savings which we can attempt to value, based on the time saved by both end users and technical staff. BT Syncordia have not assigned values to these savings, so the figures presented here are estimates made by Bloor Research, based on our experience of typical industry costs. We make the assumption that support staff and end users have an average UK salary of £25,000 pa at a total cost to the organisation of some £50,000 pa. This equates to an hourly cost of approximately £25 per person. Bloor Research believes that industry-norm availability statistics are approximately:

• 90% - conventional fat-client / thin server system
• 98% - maximum realistic fat-client / thin server system in a highly managed organisation
• 99.9% - well managed mainframe availability.

Based on information from BT Syncordia and other thin-client installations, Bloor Research believes that equivalent availability for an environment based on Citrix WinFrame and Wyse Winterm would be:

• 95% - achievable availability in an average organisation
• 99% - possible availability in a highly managed organisation.

Both of these figures represent a halving of the system down-time when compared with the equivalent fat-client model. Further, we believe that mainframe-level availability will be achievable in the near future through appropriate use of high-availability clustered servers and further improvements in management software.

Cost savings based on the productive time of end users can therefore be estimated as £2,500 per user per year for a typical environment, or £500 per user per year in a highly managed organisation. (The latter saving being over and above the saving achieved through investment in a systems management infrastructure.) These savings are the minimum values we would expect, since we have not taken into account the time needed for the user to resume productive work following the service disruption. The values assigned do not include the reduction in consequential losses mentioned above. In reality the savings will be greater than these estimates, since in the event of a single client device becoming inoperable, the user can be given a replacement unit, or be relocated rapidly to a spare desk, to further minimise the job disruption.

Statistics are not available from BT Syncordia regarding the time spent responding to support calls, nor is there a breakdown of application-related issues versus system-related issues. What is apparent from the figures available is that a much higher proportion of calls are resolved at level 1 support without the need for escalation. Even where an issue is related to an application or usability problem, WinFrame continues to increase the efficiency of the support process by providing remote session shadowing. In many cases this avoids the need to get a support engineer to the user’s workplace in order to understand the problem or demonstrate a work-around. This eliminates the totally unproductive travel time involved.

As a conservative estimate, Bloor Research believes there is an increase in efficiency of some 20% for support staff, equating to some £10,000 per support person per year. On top of these savings in responding to issues, other savings will be made due to the much more efficient management of software versioning. These savings will vary depending on the number of new releases of software to be deployed and the number of users affected. We have not attempted to quantify this as the range of values would be extremely large.

It is unlikely that these staff efficiencies will be realised by direct reductions in staffing levels. Instead users will achieve more useful work and some staff redeployment is likely. Support personnel will be able to concentrate more of their time on proactive management of the server environment to secure further increases in system availability.

It is not possible to give a definitive number of support staff in the case of BT Syncordia, since the Citrix environment currently represents a small proportion of a very large network consisting of over 1,000 servers. There is no group dedicated just to the support of the environment described above. In order to give a feeling for the total savings involved, we assume a ratio of one support person to 30 end users – a typical ratio in the industry. This gives an estimation of 14 staff dedicated to the support of the Citrix environment. We can therefore estimate the support savings to be 20% of 14 staff at £50,000 per person, or £140,000 per year.

Similarly end user savings can be estimated at 420 staff times £2,500, or £1,050,000 per year.

These total savings of £1,190,000 per year will be ongoing. Reduction in consequential losses and other intangible savings could ramp up the value to the business even further.

We would like to reiterate that these figures are estimates made by Bloor Research based on experience of a variety of different implementations, and not directly endorsed by BT Syncordia.

Future Direction

BT Syncordia will continue to deploy Citrix and Wyse solutions for their own internal systems and to offer the service to their customers. At the time of writing deployment of Microsoft Terminal Server with Citrix MetaFrame is in progress to 100 users, with a further deployment to 400 users planned.

BT Syncordia will continue to use a combination of WinFrame and Terminal Server/MetaFrame based on the requirements of individual systems. MetaFrame is not seen as a replacement technology, but as a companion product which addresses a slightly different set of needs. Initial pilots suggest that the overall costs and benefits are much the same for both WinFrame and MetaFrame.

In June 1998 Microsoft and Citrix jointly announced the availability of Microsoft Windows NT Server 4.0 Terminal Server Edition and Citrix MetaFrame. The industry had been aware of the development of these products for some time under the project names ‘Hydra’ and piCAsso’ respectively. This section provides a brief explanation of the relationship between the two products, and between them and WinFrame.

No doubt prompted by the industry acceptance of WinFrame, Microsoft identified the need for a multi-user NT environment on further versions of NT. In May 1997 Microsoft licensed MultiWin technology from Citrix and the two companies cooperated to develop the multi-user capabilities required.

To give it its full name, the ‘Microsoft Windows NT Server 4.0 Terminal Server Edition’ is a multi-user version of NT 4.0. It provides the basic functionality consisting of the multiple-user server environment which uses Microsoft’s RDP (Remote Display Protocol) to communicate with Microsoft-based clients. The client environments supported are all versions of Windows from 3.1.1 onwards plus WBTs employing the Windows CE operating system. There are appropriate extensions to the administration tools to support the management and performance monitoring of user sessions. Any application which makes standard use of the Windows API should run correctly without modification. There is an independent authentication program run by VeriTest, with information available on their web site www.veritest.com .

For existing users of WinFrame, potential reasons to change to Terminal Server will be the use of the Windows 95 style GUI, the exploitation of NT 4.0 features and the safety net of formal recognition and support of the environment by Microsoft.

So what does Citrix MetaFrame add to Terminal Server?

Most obviously, Citrix adds the ICA (Independent Computing Architecture) protocol developed for WinFrame. The ICA protocol supports a much more diverse set of client devices than RDP, including DOS, Mac OS, UNIX and Java.  The protocol also enables all popular network protocols and connections to be used, and features powerful compression which makes the execution of 32Bit Windows applications possible over low bandwidth dial-up connections. Applications can be incorporated into HTML pages to provide seamless execution from a web browser. Users with Windows PCs may execute local applications on the desktop at the same time as server-based applications, and can use standard Windows features such as cut and paste between the applications just as if they were all running locally.

For large sites, the load balancing capability of MetaFrame permits the connection of an unlimited number of clients by organising multiple servers into a ‘server farm’. This also reduces the impact of a server failure. Users connected to a server which suffers a failure will receive a request to log on and will be reassigned to one of the remaining active servers. The user profile will be retained and little work time will be lost. The overall system availability will be higher than a service based on multiple client PCs.

Sites migrating from WinFrame to MetaFrame can do so piecemeal. Any combination of servers can be accommodated running a mix of the two products.

The title of this paper ‘Windows By Other Means’ borrows its name from the best selling report ‘The Enterprise By Other Means’ by Robin Bloor. This report, written in early 1996, explored in detail the problems brought about by the fragmentation of computer power, and in particular by the anarchistic nature of personal computers. The predictions made in the report are rapidly coming true. There is now a general recognition that centralised computing is more cost-effective, simpler to manage and more likely to reflect the true needs of a business.

The Citrix thin-client architecture provides the ability to migrate Windows PC-based applications to an environment based on centralised servers. Since the move does not require any changes to applications a pilot project can be accomplished with very low risk by running a few applications for a controlled group of users. Existing skills and equipment can be retained until the solution is seen to be working and stable. At an appropriate time true thin-client devices such as the Wyse Winterm can be deployed while still retaining full function ‘fat’ PCs for the small number of users who have a genuine need. Users who also need access to non-Windows applications can use the same client hardware since it is capable of emulating several different types of terminal device.

The experience gained by sites such as BT Syncordia is that the claimed benefits really can be realised. The combination of Citrix WinFrame and Wyse Winterms has been proven to be stable, cost-effective and acceptable to users. Users of Citrix and Wyse should expect to see quantifiable saving in hardware costs, significant reductions in support costs and improved overall availability of systems. Bloor Research has estimated the ongoing savings to BT to be in the order of £1,200,000 pa in addition to the immediate hardware-related costs which BT Syncordia have estimated to be in excess of £3Million. At the same time less tangible benefits will be made possible through simple ‘hot-desking’ giving greater user mobility and flexibility and, depending on the situation, potentially reducing the total per-user infrastructure costs of an organisation. Most importantly, though probably impossible to quantify, users can target all of their energies to the business issues they are paid to deal with instead of diluting the effort through having to manage their own personal computing environment.

Other thin-client computing architectures – such as Java-based component computing - have the potential for delivering similar levels of long-term benefit, but require significant investments to be made in new applications. There is scope for these different technologies to co-exist, and there is no need for a business to make an “either/or” decision.

Bloor Research believes that the WinFrame/Winterm model of computing will continue to grow in popularity and rate of adoption, and that organisations which do adopt it will see a rapid return on investment and a more stable information processing environment.

Postscript

This paper was written on a Wyse Winterm using Citrix WinFrame and Microsoft Word prior to desktop publishing.

Back to other white papers