Archives of the TeradataForum

Message Posted: Wed, 24 Nov 2010 @ 10:58:47 GMT


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

Subj:		Re: TD Request Processing

From:		DWellman

Hi Faible,

Technically PEs work in parallel, but they're not usually described like that. Each PE runs independently of the others. A single SQL request will be parsed, optimised etc by a single PE. Each PE is a separate vproc, consisting of multiple threads/processes and so can handle multiple requests simultaneously.

As with PEs, each AMP consists of multiple threads/processes and so will handle processing for multiple SQL requests simultaneoulsy.

What that diagram is trying to show (I think) is that requests from different PEs may get sent to different AMPs or the same AMP. This really depends on the SQL, indexes etc.

- If the PE1 job is an update request using UPI, will PE2 issue the all-AMPs request, and let the AMP 1 serialize requests from both PE1 & PE2, or will PE2 wait for the PE1 request to finish before issuing the all-AMPs request?

As per my previous answer, AMP 1 will work on both requests at the same time.

Finally, just want to confirm that it is PEs (not directly AMPs) that pass SQL results back to the clients, correct? If so, this task doesn't seem to have been covered in any of the following Parsing Engine Elements:

Parser
Optimizer
Generator
Dispatcher
Session Control

Correct, all data being returned to a client application is sent by the AMPs, through the Bynet to the PEs where it is returned to the client app.

This processing is not explicitly identified within those steps, it's somewhere between the 'session control' and 'dispatcher' (more likely this last one).

Cheers,

Dave

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


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

Attachments

Library

Quick Reference