|
|
Archives of the TeradataForum
Message Posted: Mon, 09 Jun 2003 @ 13:51:19 GMT
Subj: | | Re: Host Groups |
|
From: | | David Wellman |
Thomas,
My understanding is that a Host Group defines a 'logical host' to the Teradata system. This logical host has a number of Parsing Engines
(PEs) defined to it that allow sessions to logon. The number of PEs determine the number of concurrent sessions that can be logged on from
that logical host, therefore the assignment of PEs to host groups sets one of the capacities of the Teradata system. Additionally, the
Teradata s/w implements session balancing (NOT load balancing) across all the PEs in a single host group. For a mainframe host group, this
session balancing is done by the TDP, for a lan host group this is done on the Teradata server itself (I'm guessing by the gateway
software).
Apart from the session capacity aspect mentioned above, I have also seen a performance impact on archives. At one site I worked at, we
started with too few PE's defined and archives were slow. Once we increased the number of PEs we were able to improve the archive
performance dramatically. Whilst thinking about throughput, be aware that when setting up the communications with 'the outside world' the
Teradata system has Communications Processors (CPs) and Session Processors (SPs) each if which are generically known as PEs. There are
rules and limits on how many of each type you can have and the assignment of CPs to SPs, but basically, increasing SPs allows you to have
more sessions logged on, increasing CPs (within limits) allows you to increase throughput of data transfer. In the example I quoted earlier,
increasing the number of PEs meant that the number of CPs increased and it was this lasy fact which specifically gave us improved
throughput.
Hope that helps.
Dave
| |