Archives of the TeradataForum

Message Posted: Thu, 17 Oct 2002 @ 08:13:22 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Bill,

I've seen one system set this parameter to 8 and gain clock time benefits, and that was on an old 5100 with MCA (i.e. "bus and tag") connects ! So probably increasing it above 6 will see some gains, but...

Firstly, you may want to check your configuration. How many portson the PBSA cards (i.e. the ESCON cards in the Node cabinets) do you have connected? There are 8 ports available on each PBSA card, each port corresponds to one or more PE's. I've found in the past that typically 2 or 3 connections per card gives you the max throughput when using a single TDP. Each port also corresponds to a pair of device addresses from the MVS perspective, so if you utilise more ports you need to change the MVS config.

Secondly, from a monitoring perspective I start by looking at the clock time for the major phases of the jobs. In Fastload thats Phase 1 vs. Phase 2 and Multiload that's Acquisition vs. Application. Bear in mind that changing the pbsa/escon config will only affect the data transfer phases (Phase 1/Acquisition) so if these only account for 20% of the total clock time, that's all you're going to save by making these types of changes.

Next, use the ResusageShst table to look at the throughput on the channel connects, you'll need to do some fairly simple maths on the base numbers. An IBM Escon channel can handle about 12MB/sec sustained (burst of 17MB/sec I think). If increasing the number of concurrent load jobs doesn't increase the measured throughput then you're hitting a bottle neck somewhere else (yeah I know, "state the obvious" !). Other areas to look at may be:

-- mvs utilisation, again just for data transfer phases: I've seen a maxed-out MVS system virtually halt the data loading.

-- tdp priority (linked to "mvs utilisation"): making TDP non-swap, which may even be mandatory now, and increasing it's priority can help with the data transfer phases

-- are the escon channel shared with other devices?

-- what's the speed of the devices where the input data resides?

Those are some of the things that you may care to check on. Any more questions feel free to ask.

Cheers,

Dave

Ward Analytics Ltd: Information in motion (www.ward-analytics.com)

	<Prev	Next>		<<First	<Prev	Next>	Last>>