Archives of the TeradataForum

Message Posted: Wed, 01 May 2002 @ 21:05:56 GMT

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

<-- Anonymously Posted: Wednesday, May 01, 2002 16:53 -->

We have a newly-implemented dedicated network between our Teradata box and a standalone AIX box. Both have 1GB NIC cards and are connected by fiber cabling through a high-speed switch. We plan to move about 10GB of data a week across the network in one big FastExport to the AIX box from Teradata for application processing that can't occur on the nodes.

We are performance tuning the data transfer between these two as it has not performed to expectations yet. We are trying to maximize throughput for big single-step data movements, and suspect that the TCP window size is not big enough. Does anyone have an opinion on whether we would improve throughput by increasing the following values from defaults to these values, or should we be looking at other areas?

tcp_sendspace = 521000
tcp_recvspace = 521000
udp_recvspace = 65536
udp_recvspace = 521000

Thanks in advance!

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