A Research Note by Wikibon, 2010

Executive Summary

Wikibon conducted in depth interviews with senior people who are using Unisys mainframe integrated stacks. The mainframe systems were all supporting mission critical OLTP. Wikibon found that the CapEx costs of the Unisys systems were higher than open systems, but the software development, operational costs, and maintenance costs were significantly lower. Overall the cost of a typical mainframe integrated stacks mission critical OLTP system was 19% lower at $2,383/OLTP seat/year.

The most important finding was that the availability and speed of change for the integrated stack was significantly better, leading to potential improved productivity for the end-user. About 80% of the total cost to the business are the end-user costs, and small improvements in productivity have significant effects of the overall cost efficiency of the OLTP systems.

Wikibon believes that integrated stacks will be become more common over the next decade, as solutions from VMware, Oracle, and other vendors develop. However, these solutions are developing up from the opposite end of the spectrum and are currently largely focused on small, less mission-critical systems. Wikibon believes that these systems will mature and functionally improve over the next few years and become suitable for a much wider set of applications. However, Wikibon believes that it is unlikely that these developments will either become as cost-effective or provide the same levels of security, availability, and speed of change as current mainframe systems for high-performance, mission-critical, OLTP systems until late in the decade, well beyond any strategic planning horizons.

Wikibon concludes that for performance- and security-critical OLTP systems that need very high levels of availability and business continuance, mainframes are the most cost effective solution in the marketplace. Current users of mainframes should continue to use them and expand their usage when new projects require their level of functionality.

Introduction

The first deployments of integrated stacks were found in mainframe systems, now represented by IBM and Unisys. The mainframe system integrated all the major software components into a single stack and designed the hardware to operate the software more efficiently, with particular emphasis on performance, high availability, and security. The mainframe with the least number of external software components and highest level of stack integration is the Unisys system. (The IBM mainframe typically has a larger number of external packages, especially in the management area.)

In the open systems area, integrated IT infrastructure stacks are seen as a key trend to reducing IT infrastructure costs and improving availability and security. Many IT companies are investing heavily in creating such stacks, with EMC, VMware, and Oracle currently being the most discussed and having the most aggressive roadmaps. These stacks include all the operating system, data management, backup, security, and manageability components integrated into a single package. In the case of Oracle, database software is also included. The main idea of the integrated stack is to encapsulate the application together with the metadata that defines the applications service level requirements (SLAs) and the resource requirements of the application.

Integrated stacks are projected to have a premium on initial delivery but aim to reduce the costs and risks of designing a “bespoke” stack of IT infrastructure hardware and software components by saving time, effort, and significant long-term maintenance costs. Although designing this integrated stack in-house should in theory be more efficient and initially less expensive, the complexity of designing, testing, and maintaining a homegrown integrated stack brings business risk, especially in the areas of home-grown security and business continuance. Larry Ellison of Oracle has declared his goal of becoming the “iPod of the data center” in terms of ease of deployment and overall functionality.

The senior executives on Wikibon are asking three questions about this trend:

1. Are integrated stacks cost-justified?
2. When (if ever) should mission-critical applications be migrated to this platform?
3. Where do mainframe integrated stacks fit in this landscape?

Wikibon decided to address these questions by looking at the cost justification of a mainframe integrated stack, which has been around for a long time and is is used to support many mission-critical applications. We chose to look at Unisys mainframes as the most integrated stack within the mainframe marketplace. Wikibon collaborated with Unisys customers to interview seven senior managers responsible for these Unisys integrated stacks. We wanted to know the cost-justification for continuing to use the platform, and when (if ever) they would choose to migrate to another platform.

Methodology

The methodology used was in-depth discussion with senior IT personnel in organizations with Unisys mainframes. They were chosen because they could understand the business importance of the technical decisions being made, and they were or had been part of the management team responsible for the technology decisions. The discussion guide is included in the Appendix. The discussion concentrated on the areas of knowledge and experience of the interviewees. Not all the questions areas were covered in each interview. The emphasis was on quality of information and quantification in areas where the interviewees had deep knowledge. In some of the interviews, additional information was obtained with follow-on emails.

Additional open systems performance, cost, and operational data from previous Wikibon research was used to develop a comparative costing of integrated stacks and the traditional open systems approach.

Findings

There was a surprising convergence of opinion of the users of these systems. The key findings were as follows:

(a) There was a clear strategic fit of mainframe systems. All seven participants were using the integrated stack architecture of the mainframes on mission-critical OLTP applications, with high transaction rates and high availability requirements. Six of the seven had stripped the mainframe of all other workloads. The applications were all functionally rich. One large European bank had chosen the application running on Unisys because it supported real-time data processing, rather than the traditional shadow updating and overnight reconciliation. It has become the standard platform for retail banking and accounts from bank acquisitions are migrated to this standard platform. The integrated stack enables easier implementation and maintenance of this type of very complex application than open system alternatives.

(b) Speed of change was identified by five of the seven participants as a key benefit of the integrated platform. The speed at which they could make changes, test the changes, and introduce them into the marketplace gave them a competitive edge in the marketplace.

One banking customer’s efficiency ratio was about 20 basis points compared with an industry average of 50 basis points (50 cents on the dollar). They attributed this success to two main factors:

1. A strong internal culture of parsimony within the bank – spending the money as if it is your own.
2. The speed of change possible by using an integrated Unisys mainframe and banking package as the only system to support the bank. This flexibility also allowed the rapid integration of other banks that were acquired.

(c) Availability was found to be very high by all seven of the participants. The Unisys mainframe systems almost never went down. When they did have problems, they were easy to identify and fix. Unisys provided very high levels of support for any fix that was required in the operating system or core technologies. Very few external packages were operating on the Unisys mainframes, with fewer than four applications per installation. One interviewee said: “If the mainframe system has a software problem, there are eight people in the installation, any of whom can fix the problem on their own. If an open system goes down there is an army of people, each of whom can only address a small portion of the problem.” With hardware becoming a less important contribution to overall availability, Wikibon believes that the integrated stack that was built, tested, and maintained by a single vendor is the most important contribution to high availability.

(d) Six of the seven organizations interviewed believed that the comparative operational (OpEx) costs of running the operations of the Unisys mainframe integrated stacks were significantly lower than equivalent open systems that had similar functionality and similar availability designs. One organization believed they were just a little lower. The major reasons for lower cost were:

1. The utilization of hardware was much higher. All seven companies interviewed used the Unisys metering system, which allows additional hardware to be available to handle peaks (such as month-end, year-end or seasonal changes) and paid for only as used. With a typical peak-to-average ratio of 2:1, this decreased both hardware and software costs of about 30%.
2. The support costs were much lower for the class of application (mission critical OLTP) being run on the mainframe. For some participants it was difficult to compare support with equivalent systems, but where it was possible, costs were 50% to 80% lower.

(e) There were three areas of concern about the Unisys integrated stack:

1. Four of the seven participates found the efficiency of Unisys development tools being used to be an issue. They observed that the functionality of the Unisys tools being used by their developers was below that of best-of-breed development suites available in the open systems market. However, this was not a concern for users of packages and users of the Enterprise Application Environment, (EAE, formerly called LINC) software. EAE was praised for its productivity in developing integrated OLTP/database applications. Unisys has a number of high productivity development platforms for its integrated stack, particularly in the area of integration with Web applications. Open source tools and other products such as Eclipse and Microsoft Visual Studio were being used for mainframe development. It appears that some customers are either not investing in these tools or are unaware of them.
2. Four of the seven participants said internal “sticker shock” for hardware and software expenditure on the Unisys integrated stack at renewal or upgrade time was an issue. All the participants were ultimately successful in selling the value of the integrated stack. One participant was in a government environment, where political factors were particularly important. The problem was not the overall yearly cost but a concentration of CapEx expenditure if capital budgets were under pressure; sometimes leasing was a way of ameliorating the problem.
3. Two of the participants highlighted staff availability issues. In all cases this was not a current issue, but a concern that it would be difficult to obtain experienced staff in the future as many of the current staff were due to retire soon. One potential solution mentioned during the interviews was outsourcing support to Unisys. They also mentioned the Unisys strategy of facilitating industry standard tools and skills to develop, deploy, and manage ClearPath-based applications.

(f) None of the participants found problems with the security of the mainframe platform. There were no reported problems of security breaches, in stark contrast to their experiences with open platforms.

Cost Comparisons

Wikibon took the data from the interviews plus additional data from Wikibon Research to compare the costs of the two approaches. Figure 1 shows the comparative costs of the two approaches. The key findings were that the cost of application development and maintenance was significantly lower, even taking into account the availability of better development tools on the open systems platform. The CapEx costs for the hardware and OS were significantly higher for the integrated systems, but operations and tech support significantly lower.

Figure 1 - Yearly Cost of Mission Critical High Availability OLTP Systems

Table 1 shows the yearly IT costs and the end-user costs of a typical OLTP system, and shows that the end-user cost is about four times the cost of the IT system. One of the key findings of the research was that speed-of-change of the system availability were much higher with the integrated stack. These would reflect in greater productivity benefits for the mainframe integrated stack.

Table 1 - Total End-user and IT Costs for Mission Critical OLTP Systems

Discussion

High-availability OLTP systems and the Unisys mainframe integrated platform are a clear strategic fit. The overall availability, security, ease-of-use, and speed-of-change create a great fit for high-security, revenue-facing applications that need good performance, good functionality, very high availability and speed to change and adaptability.

This solution has higher CapEx costs than would be expected in an alternate open systems solution that was configured to meet the availability, performance, and security requirements, which were to some extent mitigated by the metering option from Unisys that allowed peaks to be met with much lower operational costs. The overall cost of the Unisys mainframe, taking into account maintenance and staffing, was significantly lower than open system environments.

The flip side of this is that the higher CapEx cost of the platform meant that it was only used for applications for which it was a strategic fit. Most work that was not part of the core application had been siphoned off to other platforms.

There was very high satisfaction with the metering solution from Unisys for meeting peak requirements with much lower operational costs. Without it, hardware and OS costs would be about 30% higher.

Conclusions and Recommendations

Wikibon concludes that mainframe integrated stacks lead to higher levels of availability, security, and speed of change, which translate into improved staff productivity. While the CapEx costs of hardware and OS were higher for the integrated stack, the overall cost for integrated stacks were 19% lower when software development and operational support costs were included.

The integrated stacks currently available on open systems fall far short of the functionality available on mainframes. This is particularly true in the security and systems management areas. Wikibon reviewed the security available on open platform virtualized stacks with leading security experts and concluded that the only way to to ensure that an application could not compromise other applications was to run each application on a physically separate machine. There are industry moves to make open system hardware architectural improvements to enable this, but will take a very long time before security is integrated into hardware, hypervisors, operating systems, and middleware with the same trusted system security that is available now on mainframes.

In general, Wikibon expects future open system integrated stacks to move towards the functionality available in the mainframe integrated stacks. There will be no free lunch: as with the mainframe, the open system integrated stacks of the future will have higher CapEx costs, and will drive lower software development and operational costs. The key benefits of the integrated stacks will be faster time to deploy and update, lower costs of support, and (eventually) higher levels of availability and security.

Wikibon believes that integrated stacks will become more common over the next decade, as solutions from VMware, Oracle, and other vendors develop and mature. However, these solutions are developing up from the opposite end of the spectrum and are currently largely focused on small, less mission-critical systems. Wikibon believes that it is unlikely that these developments will either become as cost effective or provide the same levels of security, availability, and speed-of-change as current mainframe systems for high-performance, mission-critical OLTP systems until late in this decade, and well beyond any strategic planning horizons.

Wikibon concludes that for performance-and security-critical OLTP systems that need very high levels of availability and business continuance, mainframes provide the most cost-effective solution in the marketplace. Current users of mainframes should continue to use them and expand their usage when new projects require that functionality.

Appendix: Strategic Fit Discussion Guide

Interviewee Information

Name:

Title:

Company:

Tel Number:

Email:

Date of Interview:

Discussion Scope

The scope of the discussion was to explore the business value and strategic fit of mainframe and open systems.

The key issues to explore are about the value of application systems based on integrated mainframes and the business value they contribute to the organization. This include the strategic fit of applications to Unisys mainframes, and the strategic fit of other mainframe and open systems in the organization. The discussion should lead to a high level view of the resources used for different systems, and the value generated.

Detailed Interview

The interview is in three parts:

1. A discussion about the Unisys mainframe systems they have installed, the applications run and the support environment
2. A discussion of other mainframe and other systems
3. A discussion about value and strategic fit

Overall Business Metrics

Some questions about the demographics of the organization

• • Number of WW employees
  • Primary business, industry or organizational objectives
  • Total organization wide IT budget for 2009
  • Total revenue (budget for non-revenue organizations)
  • Primary metric that the company uses to determine success, e.g. revenue, market share, sales, etc
  • Total WW value of the company success metric
  • Name of Division/Divisions within the company supported by the interviewees IT department
  • Number of employees in the division

IT Applications & Infrastructure

Major System technologies that support the business

Unisys Systems

• • Major Systems Supported
  • Revenue or budget base supported
  • Users supported
  • Availability required
  • Importance/value of system to the organization
  • Budget/resources applied to system (CapEx & OpEx)
  • Relationship of system to other systems

Other Mainframe

• • Major Systems Supported
  • Revenue or budget base supported
  • Users supported
  • Availability required
  • Importance/value of system to the organization
  • Budget/resources applied to system (CapEx & OpEx)
  • Relationship of system to other systems

Open Systems

• • Major Systems Supported
  • Revenue or budget base supported
  • Users supported
  • Availability required
  • Importance/value of system to the organization
  • Budget/resources applied to system (CapEx & OpEx)
  • Relationship of system to other systems

Strategic Fit of Systems

The premise of strategic fit was explained (if necessary) and discussed. The key questions that were asked were what types of application were supported on each platform, and what the criteria were for choosing the infrastructure type.

How are decisions made on the strategic fit of IT Platforms and Applications?

• • Factors taken into consideration
  • Cost expectations

Advice to other users

Any advice either general or specific that could be passed on to other users facing

Other feedback

A general catchall question to ensure that there was no important information that the interviewer or interviewee wanted to ensure was included.

Footnotes: 1 Efficiency ratio is a technical banking term. The ABA Banking Journal characterizes efficiency ratio as follows: ‘The efficiency ratio is a useful tool in determining how effectively a bank or thrift is generating revenue, and as one would expect there are many different ways to calculate the number. Typically one thinks of the ratio as general and administrative expense as a percentage of fully taxable equivalent revenue: the lower the number the more efficient the company.’ (See: http://www.allbusiness.com/finance-insurance/credit-intermediation-related-activities/431012-1.html)

Action Item:

Footnotes: