home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cis.ohio-state.edu!ucbvax!ludwig.sharebase.com!glenn
- From: glenn@ludwig.sharebase.com (Glenn Linderman)
- Newsgroups: comp.databases
- Subject: Re: Hot Standby DBMS's
- Message-ID: <9209091721.AA27050@ludwig.sharebase.com>
- Date: 9 Sep 92 17:21:01 GMT
- Sender: daemon@ucbvax.BERKELEY.EDU
- Lines: 52
-
- -> From: dhepner@cup.hp.com (Dan Hepner)
- -> Subject: Re: Hot Standby DBMS's
-
- -> From: dboyce@hemel.bull.co.uk (David Boyce)
-
- -> >Interfaces for obtaining journal data from an RM and passing it to an
- -> >standby RM would be needed, as would RM-independent methods of recovery
- -> >and catch-up. Not beyond the wit of man, I would have thought.
- -> >
- -> >Obviously the database vendors need to be convinced of the need to make
- -> >this interface available.
-
-
- -> Any ideas on how to convince them? In particular, it needs to be
- -> shown what can be done using this scheme which the DB vendors
- -> could not do themselves. Intuitively, one would say "but it's
- -> more open". In what way does this openness benefit anyone?
-
- -> Dan Hepner
-
- The point of passing the journal data from the master RM to a Hot Standby RM
- is performance: it generally takes less work (read less time) to process
- journal entries than it does to process the original requests.
-
- Let's take a look at why: the journal is a log of physical updates to the
- database pages (if it isn't, it probably also isn't faster than the
- original requests). Being physical, however, it is related to the particular
- (proprietary) data organization used by the DBMS.
-
- Translating things like block size and (index and data) record formats
- from one proprietary form to another, or even to a canonical form, to pass
- to another DBMS via a TM doesn't sound like fun, if it is even possible.
- It also sounds very error prone to me, since the combinatorics would make
- testing very difficult.
-
- As a result, then, you would have an extremely complicated translation
- system that would probably be as slow as or slower than the original
- queries. So passing the original queries to both machines in parallel
- via a TM sounds like a pretty good alternative, and you can have that
- today.
-
- The alternative is to rewrite all DBMSs to have the same internal
- structures and logging techniques. Would you suggest the DBase format
- as a standard that everyone knows??? :-) <- that's a joke, please, but
- it does make a point, even if DBase doesn't do logging...
-
- To play the catchup game would require that the surviving DBMS or the TM
- keep a record of the logical transactions processed while the other DBMS(s)
- failed, so they can be replayed later to catch up. If the load is light
- enough that they can catch up by applying logical transactions...
-
- GLenn
-