This is the 10th excerpt from the second book in the Defen series: BIT: Business Information Technology: Foundations, Infrastructure, and Culture
Note that the section this is taken from, on the evolution of the data processing culture, includes numerous illustrations and note tables omitted here.
--- This is a virtual tour of a data processing center. There are several important things to note about this:
Scale and Budget
This is a typical scale data center for an international insurance underwriter serving the head office organization and just over 1,600 sales offices in the United States and Canada.
Neither the Director of Information Systems nor the Finance VP he reports to will reveal the current information technology budget or share a copy of the service level agreement. All values used here are therefore estimates. Reliable external estimates suggest that fiscal 2005/6 expenditures for this center ran to about:
|Estimates are for 2005/6||Employees||Budget (Millions of USD)|
|Data center capital lease management||4||18|
|Data Center Operations||520||60|
|Business Desktop Services||38||80|
Data communications and user desktop maintenance and support are contracted to IBM Global Services.
Until late 1998 all offices used IBM 3278 terminals and RJE (remote job entry) on 3274 style controllers to connect to mainframe services for the two most mission critical applications clusters: Customer Claims (query and processing) and Sales Management (order and quotation processing). At that time the company rolled out new IBM Netvista desktops to all offices and instituted a seven year agreement, since extended by two years, with IBM Global Services to provision, manage, and support the company's 15,000 desktops along with just over 1,800 Windows 2003/XP servers located outside the data center.
The original Netvista PCs had SNA boards enabling the machines to emulate a 3278 style terminal. In the most recent round of desktop upgrades these boards were not replaced because the latest generation mainframes now use TCP/IP (the internet standard networking protocol) and 327X emulation can therefore be handled entirely in software. Most applications requiring RJE or a 3279 terminal have, furthermore, had customized Windows 2000 clients, since updated to 2003/XP, developed to replace the old character interfaces.
The Windows 2003/XP servers located outside the data center handle Microsoft Exchange, Lotus Notes (now IBM Domino), and departmental file and print sharing.
The data center itself started with a System 370 in 1973, was most recently rebuilt to house four S/390 processors in 1997, and recently transitioned to a pair of IBM zSeries 2064-216 mainframes in a two way sysplex (clustered) configuration.
The five year capital lease on the new IBM mainframe gear includes:
$876,713 monthly in respect of the base hardware (the list price of the gear as delivered is secret, but estimated at around $51 million).
$138,000 monthly in service and hardware support fees including support for the tape automation brought forward from the S/390 environment and four new SHARK/ESS two terabyte data stores;
$128,400 per month (rising to $162,300 in 2006/7 when software fee caps agreed to as part of the transition expire) in software usage fees;
$190,220 monthly as the capitalized value of the transition services contract under which 33 IBM Global Services staff helped transition applications code from the S/390 to the current zSeries 64bit environment.
--- Two footnotes from the text:
Page mode terminals transmitted and received information one page at a time, not one character or line at a time. Thus users typically filled out an on-screen form, and then sent the whole form -a bit like a page submit on a web form- for validation and processing.
"Sysplex" is the IBM name for a group of two or more tightly linked machines in which an automated job scheduler can allocate work to either machine depending on available resources and one machine can therefore take over some jobs for another if that unit fails.
In the mainframe world this type of clustering - and be aware that the word "clustering" takes on quite different meanings for different technology groups- means both failover and resource sharing.
As such operation usually requires shared external disk arrays such as IBM's SHARK/ESS data stores and a dedicated 1GBS I/O connection between each pair of machines.
Notice that getting the facts right is particularly important for BIT - and that the length of the thing plus the complexity of the terminology and ideas introduced suggest that any explanatory anecdotes anyone may want to contribute could be valuable.