Pat O'Keefe writes: >The improved SLA on the back-end is anything but "free". The >improved SLAs come only when the applications are redesigned to >take advantage of the improved availability of the underlying >platforms. Multiple applications that share a redesigned >component may get some shared improvements in SLA but only >when the shared component is the SLA's problem child.
OK, but there are many occasions, probably particularly on mainframes, when the shared component(s) is(are) the problem child(ren). You're saying there are exceptions -- sure, that's true. But.... >Adding continuous availability to large applications (or systems >of applications) that weren't designed around the concept takes >a lot of time and effort (read "money"). And a lot of banks aren't >too free with their money right now. An existing solution (even >when based on antiquated technology) has strong appeal in the >current financial climate. ....It's all comparative. Highly available front-ends are not cheap, and they only offer degrade mode customer service anyhow. As a *general rule*, it is better to spend money once (in one logical place) on highly available infrastructure to deliver high SLAs rather than spend in multiple channels. If reasonably possible, which it usually is. You're saying there are exceptions. Yes, true, no disagreement. You can also smoke tobacco and live past 100. :-) But I really, really don't think this is any sort of radical notion. Let's look at GDPS as one small example. Once an organization implements GDPS, how much does it cost to protect one more 3390-XX volume? Almost nothing is the correct answer, in general. Once you configure one MQ shared queue in the Coupling Facility, how much to configure the second one? Also almost nothing, in general. And so on. It just makes tons of economic sense -- IN GENERAL -- to establish the highly available infrastructure to deliver high SLAs once, then share it as much as possible. To pick an analogy, if you ran a hospital, would it make economic sense to buy standby electrical systems (such as UPS batteries) for each and every piece of critical medical equipment, or would it make better sense to buy one backup generator for the entire hospital? Unless you're a very small hospital, of course it's the latter. (And yes, you might have some rewiring to do in the hospital, but it still makes perfect sense.) And once you've got the backup generator, the cost for that backup service when you plug in another heart monitor is negligible. This is also reason #684 why crafting an accurate and meaningful chargeback regime is hard. Sharing makes overall economic sense to the organization, but imagine levying chargebacks equally for HA services across all users. Users vary in how much they value such services, and the costs of delivering those services are dramatically different after the first user. In theory a rational administrator could charge a much lower rate to certain users *and* collect more total money to fund shared HA services. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific E-Mail: [EMAIL PROTECTED] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
