Commentaries
Comments and observations by Paul Strassmann
Sunday, March 13, 2005
About Information Architectures – Part I – Telephony and Mainframe Computers
Where you invest in information technologies to endow a complex system with intelligence is perhaps the most important single decision to make when defining the “architecture” of how hardware, software and communications will work.
The winning architecture for telephony was conceived in the 1880’s. It offered voice communications with maximum economy by managing complexity. Voice was transmitted as an analogue signal through very intelligent and expensive central switches that remained ignorant of the contents of the messages. This made it possible to use very inexpensive and “dumb” terminal equipment. For the customer neither labor costs nor special skills were needed. The dominant phone company – a legalized monopoly - incurred all of the operating and capital costs. They reaped enormous profits without much risk because they could control the introduction of innovation. The customer ended paying “rent” of the service through mostly variable charges.
The telephony architecture reached its limits of growth when the cost of capital for extending the capabilities of the central switches for serving the rising complexity of demands for improved services became prohibitive. The previously successful telephony solution ran out of the economic as well as the technological capacity to earn profits.
SUMMARY OF THE PHONE ARCHITECTURE: 1. Vendor incurs huge capital costs, moderate operating costs. 2. Customer has no capital costs, zero operating costs. 3. Limits on of the architecture reached when further additions to the vendor’s capital costs are ineffective and not profitable.
------------------------------------
The winning architecture for computer mainframe-based systems was conceived in the 1960’s. It solved the problem of delivering computer services while mastering complexity with maximum efficiency. Digitized data was passed to incredibly smart and expensive central computers combined with expensive intelligent switches to handle communications. That made it possible to use relatively inexpensive “dumb” terminals to serve customers. Central computers were programmed to recognize the format of the data for passing to application software. It required only a moderate amount of training for a customer to use a computer terminal. Meanwhile corporations had to pay for a steadily increasing cadre of specialists to keep the systems operational. Vendors sold the computer equipment as capital goods and also charged for technical support. One computer firm emerged as a quasi-monopoly for a brief period. It reaped huge but risky profits because it was unable to completely control innovation.
The customer now incurred the entire cost of capital as well as all operating expenses. The total costs of computerization where largely fixed, with little variability except for a steady escalation in expenses to keep up with rising demands.
The mainframe computer architecture reached its limits of growth when the cost of capital for extending the capabilities of the central mainframe computes for serving the rising complexity and proliferation of demands for improved services by customers became prohibitive. The previously successful mainframe computer solution ran out of the economic as well as the technological capacity to increase the dominant vendor’s profits.
SUMMARY OF THE MAINFRAME COMPUTER ARCHITECTURE: 1. Vendor passes capital costs to customers while maintaining moderate operating costs. 2. Customer incurs rising capital costs and large operating costs. 3. Limits on of the architecture reached when further addition to the customer’s capital costs are ineffective and not profitable.
CONCLUSION: The choice of an "architecture of information" is primarily one of economics and only secondarily as enabled by changes in the prices of technolgy. The single most important design issue is in answering the question: Who pays? Successful (and temporarily sustainable information architectures) are the result of an unstable balance between what a customer can afford to manage and what the technology choices can deliver.
The winning architecture for telephony was conceived in the 1880’s. It offered voice communications with maximum economy by managing complexity. Voice was transmitted as an analogue signal through very intelligent and expensive central switches that remained ignorant of the contents of the messages. This made it possible to use very inexpensive and “dumb” terminal equipment. For the customer neither labor costs nor special skills were needed. The dominant phone company – a legalized monopoly - incurred all of the operating and capital costs. They reaped enormous profits without much risk because they could control the introduction of innovation. The customer ended paying “rent” of the service through mostly variable charges.
The telephony architecture reached its limits of growth when the cost of capital for extending the capabilities of the central switches for serving the rising complexity of demands for improved services became prohibitive. The previously successful telephony solution ran out of the economic as well as the technological capacity to earn profits.
SUMMARY OF THE PHONE ARCHITECTURE: 1. Vendor incurs huge capital costs, moderate operating costs. 2. Customer has no capital costs, zero operating costs. 3. Limits on of the architecture reached when further additions to the vendor’s capital costs are ineffective and not profitable.
------------------------------------
The winning architecture for computer mainframe-based systems was conceived in the 1960’s. It solved the problem of delivering computer services while mastering complexity with maximum efficiency. Digitized data was passed to incredibly smart and expensive central computers combined with expensive intelligent switches to handle communications. That made it possible to use relatively inexpensive “dumb” terminals to serve customers. Central computers were programmed to recognize the format of the data for passing to application software. It required only a moderate amount of training for a customer to use a computer terminal. Meanwhile corporations had to pay for a steadily increasing cadre of specialists to keep the systems operational. Vendors sold the computer equipment as capital goods and also charged for technical support. One computer firm emerged as a quasi-monopoly for a brief period. It reaped huge but risky profits because it was unable to completely control innovation.
The customer now incurred the entire cost of capital as well as all operating expenses. The total costs of computerization where largely fixed, with little variability except for a steady escalation in expenses to keep up with rising demands.
The mainframe computer architecture reached its limits of growth when the cost of capital for extending the capabilities of the central mainframe computes for serving the rising complexity and proliferation of demands for improved services by customers became prohibitive. The previously successful mainframe computer solution ran out of the economic as well as the technological capacity to increase the dominant vendor’s profits.
SUMMARY OF THE MAINFRAME COMPUTER ARCHITECTURE: 1. Vendor passes capital costs to customers while maintaining moderate operating costs. 2. Customer incurs rising capital costs and large operating costs. 3. Limits on of the architecture reached when further addition to the customer’s capital costs are ineffective and not profitable.
CONCLUSION: The choice of an "architecture of information" is primarily one of economics and only secondarily as enabled by changes in the prices of technolgy. The single most important design issue is in answering the question: Who pays? Successful (and temporarily sustainable information architectures) are the result of an unstable balance between what a customer can afford to manage and what the technology choices can deliver.
Comments:
<< Home
Mr. Strassman:
I find your analysis of mainframes to be unfair and inappropriate. You do not account for the following:
!. Were it not for IBM and mainframe ISV distortive/extortive software pricing schemes, there would not be a single question in any sane person's mind what architecture they would use for business transaction processing, assuming they were not afflicted with the mindset of computer science.
2. The fundamental architecture of the mainframe recognizes important characteristics of transaction-based workloads, and delivers performance which cannot be matched on a rational comparison basis. Google might be a nice example for certain compute-intense workloads, but that is not representative of the majority of a business' computing work.
3. The discipline and reliability of mainframe environments is achievable at a lower cost than any distributed environment, even with tbe IBM/ISV distortive/extortive software pricing schemes considered once the workload size reaches a critical point.
4. Distributed systems constantly work to ape the administrative structures, processes, and efficiencies of a mainframe without much luck. If someone thinks ITIL and SNMP are strokes of brilliance, I must wonder about their real experience in computing since large networks of mainframe-based systems were more successfully managed for years than these johnny-come-lately constructs would imply.
Post a Comment
I find your analysis of mainframes to be unfair and inappropriate. You do not account for the following:
!. Were it not for IBM and mainframe ISV distortive/extortive software pricing schemes, there would not be a single question in any sane person's mind what architecture they would use for business transaction processing, assuming they were not afflicted with the mindset of computer science.
2. The fundamental architecture of the mainframe recognizes important characteristics of transaction-based workloads, and delivers performance which cannot be matched on a rational comparison basis. Google might be a nice example for certain compute-intense workloads, but that is not representative of the majority of a business' computing work.
3. The discipline and reliability of mainframe environments is achievable at a lower cost than any distributed environment, even with tbe IBM/ISV distortive/extortive software pricing schemes considered once the workload size reaches a critical point.
4. Distributed systems constantly work to ape the administrative structures, processes, and efficiencies of a mainframe without much luck. If someone thinks ITIL and SNMP are strokes of brilliance, I must wonder about their real experience in computing since large networks of mainframe-based systems were more successfully managed for years than these johnny-come-lately constructs would imply.
<< Home