3 ms·

At least that is a reasonable, sane explanation. Otherwise, the idea of encoding data in JSON and then transforming that to XML seems a little ... convoluted.

by krylon 12y ago

gaius 12y ago
It's particularly relevant on IBM as they have hardware-accelerated XML processing http://www-03.ibm.com/systems/z/hardware/features/zaap/ http://www-03.ibm.com/systems/z/hardware/features/zaap/
- krylon 12y ago
  Man, I would so love to have a mainframe. Maybe, one day... UPDATE: I think I am getting the picture. This is about interfacing legacy systems to new ones. In that context, it makes perfect sense.
- mrsteveman1 12y ago
  From what I read about zAAP and zIIP, they're not actually hardware accelerators like a DSP or a purpose built processor, rather they're a way to reduce the cost of running specific workloads on a mainframe, and IBM markets them that way[1]. The zIIPs themselves seem to be standard z processing units that have been loaded with microcode that prevents them from doing anything other than run the workloads they were purchased for[2]. Adding (or activating/licensing) a general processing unit instead would get you the same "acceleration", but it would cost more due to the way IBM structures mainframe pricing. They know you can't use these zIIP licensed processors for general processing tasks, so they don't charge you as much for them upfront, and they don't count for software licensing charges in other ways. Nothing quite like this exists in the x86 world, but it does remind me a little of the time when AMD/Intel were selling 1/2/3 core CPUs that technically had 2/4 cores in the package, with the extras intentionally disabled to create CPU models for a specific market price point, or because they didn't work properly when tested at the fab. I still have a motherboard here that can enable the disabled cores on certain AMD processors (with a warning in the BIOS that they may not actually work). [1] http://ibmdatamag.com/2013/04/performance-and-capacity-considerations-for-ziip-engines/ http://ibmdatamag.com/2013/04/performance-and-capacity-consi... [2] http://www.itjungle.com/big/big013106-story01.html http://www.itjungle.com/big/big013106-story01.html

At least that is a reasonable, sane explanation. Otherwise, the idea of encoding data in JSON and then transforming that to XML seems a little ... convoluted.

by krylon 12y ago

At least that is a reasonable, sane explanation. Otherwise, the idea of encoding data in JSON and then transforming that to XML seems a little ... convoluted.

gaius 12y ago

It's particularly relevant on IBM as they have hardware-accelerated XML processing http://www-03.ibm.com/systems/z/hardware/features/zaap/ http://www-03.ibm.com/systems/z/hardware/features/zaap/

krylon 12y ago
Man, I would so love to have a mainframe. Maybe, one day... UPDATE: I think I am getting the picture. This is about interfacing legacy systems to new ones. In that context, it makes perfect sense.
mrsteveman1 12y ago
From what I read about zAAP and zIIP, they're not actually hardware accelerators like a DSP or a purpose built processor, rather they're a way to reduce the cost of running specific workloads on a mainframe, and IBM markets them that way[1]. The zIIPs themselves seem to be standard z processing units that have been loaded with microcode that prevents them from doing anything other than run the workloads they were purchased for[2]. Adding (or activating/licensing) a general processing unit instead would get you the same "acceleration", but it would cost more due to the way IBM structures mainframe pricing. They know you can't use these zIIP licensed processors for general processing tasks, so they don't charge you as much for them upfront, and they don't count for software licensing charges in other ways. Nothing quite like this exists in the x86 world, but it does remind me a little of the time when AMD/Intel were selling 1/2/3 core CPUs that technically had 2/4 cores in the package, with the extras intentionally disabled to create CPU models for a specific market price point, or because they didn't work properly when tested at the fab. I still have a motherboard here that can enable the disabled cores on certain AMD processors (with a warning in the BIOS that they may not actually work). [1] http://ibmdatamag.com/2013/04/performance-and-capacity-considerations-for-ziip-engines/ http://ibmdatamag.com/2013/04/performance-and-capacity-consi... [2] http://www.itjungle.com/big/big013106-story01.html http://www.itjungle.com/big/big013106-story01.html