Sybase IQ Questions

1) Every write server (if there is more than one) is in charge of distributing its writes through Table Level Versioning (TLV) Log. This lets the secondary servers know about table versions. Remember IQ effectively locks a table snapshot (although it doesn't copy the whole table, just the changed blocks). TLV in Multiplex is just like the Lock Manager in any traditional RDBMS. You get one of these per writer node.

2) The data is the FP index. There is always an FP on every column because that's where the column's data lives. The ROWID on every column is just the array entry. Remember IQ stores everything in columns, there is actually no real "row" anywhere. "Row #10" just means "the 10th entry in the FP array of column1 + the 10th entry in the FP array of column2" and so on. Is that making sense? The ROWID is not stored anywhere, just as the number of each array entry is also not stored anywhere, it is the data structure.

3) Every index is rebuilt during loads. Unlike traditional RDBMS, it is faster to index an empty table then load it than the other way around. Test it and see. :)

Thanks Joe.

Whenever the failed secondary nodes comes back online, there is an internal process which will read TLV log from IQ store and sync the catalog store of that secondary node. Am I right?

Now I understand the ROWID concept in FP index . What about in LF index, is the ROWID attached to LF index column values?

Regards
CS

Thanks Joe.

Regards
CS