Archives of the TeradataForum

Message Posted: Mon, 06 Jan 2003 @ 18:35:46 GMT

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

<-- Anonymously Posted: Monday, January 06, 2003 13:34 -->

I understand the basic concept of fallback and RAID. What I am trying to assess is the marginal benefit of using fallback when RAID is already employed. What types of failures would cause an outage or loss of access to data if we did not use fallback? Let me start by attempting to categorize the scenarios. Let's assume a configuration of 8 nodes with 2 four node cliques. I understand that some of the answers might depend on the cabling, power supply routing, or other issues. Are the scenarios under A and B correct? What about the scenarios in C? Are there other scenarios that are commonly encountered?

A. When it does NOT help:

1. Single disk failure

2. Multiple disk failures assuming only one failure per rank.

3. CPU outage - system keeps running without disruption on less CPUs

B. When it provides value:

1.) Multiple disk failures in the same rank.

C. Not sure

1.) Power outage to one supply line, assume a second power supply line feeds other nodes/cabinets

2.) One disk cabinet outage

3.) One processor cabinet outage

4.) Loss of single disk array controller (when only one is configured per rank)

5.) Single VAMP or VPE outage

6.) Complete node failure - vproc's will migrate to other nodes but are there common cases when this does not provide complete failover?

Any comments will be much appreciated.

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