Re: Timestamp Ordering granularity

Wojtek Narczynski wrote:
> I have a question regarding concurrency control.
>
> Locking techniques can use different granularity : table-, page-,
> record- level locks, depending on the level of contention.
>
> Can the same be made work for TO? Would it require to always store
> timestamps on each record and page?

If you use TO and assign a single timestamp to a page, then the system will treat it as though any transaction that modifies any record on the page actually modified every record on the page. That's a conservative but safe way of dealing with it.

-- 
Jonathan Leffler                   #include <disclaimer.h>
Email: jleffler_at_earthlink.net, jleffler_at_us.ibm.com
Guardian of DBD::Informix v2003.04 -- http://dbi.perl.org/