Subject | Re: [firebird-support] Re: i need your opinion... |
---|---|
Author | Ann W. Harrison |
Post date | 2005-10-29T16:27:42Z |
Adam wrote:
will stop the system. For example, most processes have a read lock on
the most recent TIP. Unless they can release that lock, it will keep
other transactions from committing. A read lock on the header page
will keep other transactions from starting. A read lock on an index
page stops inserts and updates.
Let me reiterate, first, be sure you understand the problem. In my
experience, Firebird tends to be CPU bound more often than I/O bound.
That's another issue and one that we are addressing - Arno took a
great step in that direction with the index changes - but for 1.5,
CPU is a more critical resource than disk.
Regards,
Ann
>Don't do that! If the sleeping process is holding a critical lock it
> To answer your direct question, I suppose you could write a UDF that
> sleeps the thread for a set time rather than waste clock cycles which
> may be important in sorting etc.
will stop the system. For example, most processes have a read lock on
the most recent TIP. Unless they can release that lock, it will keep
other transactions from committing. A read lock on the header page
will keep other transactions from starting. A read lock on an index
page stops inserts and updates.
Let me reiterate, first, be sure you understand the problem. In my
experience, Firebird tends to be CPU bound more often than I/O bound.
That's another issue and one that we are addressing - Arno took a
great step in that direction with the index changes - but for 1.5,
CPU is a more critical resource than disk.
Regards,
Ann