Archives of the TeradataForum

Message Posted: Mon, 07 Nov 2005 @ 14:35:52 GMT

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

<-- Anonymously Posted: Monday, November 07, 2005 09:32 -->

Thanks for replies thus far. Here are some comments in response.

Michael McBride: "Some questions about your backup process that wasn't evident in your initial posts: Are you running single stream or cluster (multi-stream)? Is your data skewed on your larger tables?"

Response: Backups are all-amps, i.e. not clustered. No data skew on larger tables. Yours sounds like a Unix backup, whereas ours is mainframe. Certainly cluster backups can probably be done faster, if cluster groups are done in parallel. Cluster restores still have to be serial, though, if I understand properly.

Fiber is much faster than ESCON from what I've heard, but I don't know how much faster.

Anomy.Anom: 30GB/hour is roughly the throughput capacity of a single ESCON connection. TDP and Teradata can use multiple connections in parallel if available, but I/O to a single mainframe "flat file" (e.g. the output from ARCMAIN) can't use more than one. Running more jobs in parallel helps until you saturate something else (e.g. number of Teradata ESCON adapters).

Response: I thought Resusage (via dbc.resusageshst) might be indicating we were using multiple ESCON connections for a single backup. I must be misinterpreting the data somehow.

- How does one determine from Resusage (or anything else) that only a single ESCON connection is in use?

- How much faster than ESCON might a fiber connection be?

Anomy.

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