>>>>> "Simon" == Simon J Mudd <simon.mudd@stripped> writes:
Simon> On Wed, 17 Mar 1999, Ed Carp wrote:
Simon> [my stuff snipped]
>> A lot of the stuff I've seen uses some sort of push technology. This can be
>> as simple as doing a select to get the rows that have changed, then push the
>> data out to the clients. Easiest way, if you have TCP/IP, is to have a
>> socket listening on the client - when the data changes, the server opens a
>> socket to the client, pushed out the data, then closes the socket. Of
>> course, you have to contend with clients having to register their IP address
>> with you and dealing with clients that go away and don't de-register. You
>> get the idea. :)
Simon> Yes, most of the PCs in the office don't use databases, and employ
Simon> proprietary technology to push the data. I don't need this performance
Simon> (up to hundreds of updates/sec, but am concerned about doing multiple
Simon> repeated queries, even if I include timestamp information to reduce the
Simon> data returned by the SELECT), as it will almost certainly produce a heavy
Simon> load on the db server).
Simon> Maybe SQL just doesn't answer this question.
Note that in most cases with SQL the performance is depending on how
many new disk blocks you are accessing.
If you have an timestamp on each row and only do selects based on
recent timestamps, MySQL will only access the same disk buffers over
and over again. Most OS cache this very good and you have to do real
many queries to get any significant load.
In MySQL 3.23 we will extend
to include a lot of information about each table !
We will at least include:
- Number of rows
- Number of deleted blocks
- Size of data file
- Size of index file
- Active keys
- Create time
- Last changed !
- Last checked
I assume the above will help you with your problem!