Re: INITRANS and MAXTRANS are confusing me?

From: Pablo Sanchez <pablo_at_dev.null>
Date: Tue, 11 Mar 2003 18:34:49 -0600
Message-ID: <Xns933BB2D622ED3pingottpingottbah@216.166.71.233>

DA Morgan <damorgan_at_exesolutions.com> wrote in news:3E6E6CAB.C34F256F_at_exesolutions.com:

> Pablo Sanchez wrote:
> 

>> "Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in
>> news:pan.2003.03.11.20.20.52.379490_at_yahoo.com.au:
>>
>> > Which is fair enough, but that is why Informix (like many other
>> > databases) do lock escalation. They see locks as a scarce resource
>> > because there is a central pool of locks which have to be divided
>> > amongst all possible users. So instead of letting you take 10000
>> > row locks, I will escalate that to a single table lock, and to hell
>> > with the consequences (like the minor one that you've just locked
>> > 5000 rows that you weren't intending to touch, but I was... so now I
>> > have to wait).
>>

> 
> True if anyone was foolish enough to SELECT-FOR-UPDATE on a large
> number of records.

The point that you miss is that no one on the in-memory lock manager world does it either. They come in, do the work and out they go.

Happens to be that the in-memory lock manager is 10x faster than a disk space lock manager (Oracle).

Also, locking escalation is configurable by the DBA. Not a biggie.

-- 
Pablo Sanchez, High-Performance Database Engineering
http://www.hpdbe.com

Received on Tue Mar 11 2003 - 18:34:49 CST