Re: MaxDB 7.5 - Some technical questions

Am Donnerstag, 27. Januar 2005 10:46 schrieb Thomas Steinmaurer:
> 1) Full-Text search
>
> Does MaxDB have a built-in full-text search facility?

No. 

> 2) Clustering, Load-Balancing
>
> I've read that MaxDB supports clustering for fail-over (hot-standby).
> Does that also support load-balancing in an clustering environment that
> means spread client request across different nodes in a cluster?

As far as I know: no. MaxDB scales by hardware. I've seen figures from 2003 
saying that MaxDB can handle 5.500 concurrent users on a 4 CPU box. That's a 
lot and MaxDB has been improved since '03 (7.3 => 7.6)...

>
> 4) Prepared Statements
>
> Does MaxDB support prepared statements that means an SQL statement gets
> "prepared" (the syntax of the SQL statement is validated, the statement
> gets compiled into a language the server understands) and executed only
> with different parameter values?

Yes. Check also the Shared SQL documentation. Here's a mailing list posting on 
it:
http://archives.neohapsis.com/archives/dev/sapdb/2003-q4/0613.html

> 5) Parts of one SQL statement in parallel
>
> In an SMP environment. Is it possible to split one SQL statement into
> parts which are executed on different CPUs?

No. 

But the process design allows the read()/write() routines to run on one CPU 
(in one thread) and the interpretation of the results, the SQL execution 
itself on a second CPU (in a different thread). 

The SQL processing is done within user kernel threads (UKT). UKTs are OS 
threads. Each UKT contains several user tasks (UT) which handle the 
individual SQL queries. Starting from 7.4.03 UTs can be switched between 
UKTs. UKTs cause the main CPU load. If a SQL query within an UT eats up all 
available CPU, than another UT on the same UKT can be switched to another UKT 
that will likely be scheduled to a different CPU by the OS. This does not 
increase the speed of an individual query but it improves the throughput 
which is often more important.

Another advantage of this process model: UKT cause the main CPU load. By 
limiting the number of UKTs you can "limit" the number of CPUs used by the 
database. Guess you have 4 CPUs and 1 UKT in 1 OS thread. This 1 UKT will be 
bound by the OS to CPU 1. CPUs 2-4 take the other MaxDB threads, dealing 
mostly with I/O, not causing much CPU load. This way MaxDB occupies only one 
CPU - as a rule of thumb. 

> 7) Memory usage > 2GB
>
> I've read MaxDB (32-bit versions) can use > 2GB memory on Windows by
> enabling AWE support. Is it possible to address more than 2GB memory on
> other platforms as well?

That's a kernel/OS issue not an application issue. If you have an OS that can 
be configure to split the 4GB adressable memory not in 2/2 pieces for the 
kernel and the userspace but in 1/3 pieces for the kernel and the userspace, 
then it's possible to use more RAM than 2GB within a process.

Check kerneltraps or similar resource for Linux and the OS documentation of 
the desired platform.

> What about the 64-bit versions of MaxDB?

Go for it. Check the manual on supported hardware platforms.

Nixnutz

-- 
MaxDB Discussion Mailing List
For list archives: http://lists.mysql.com/maxdb
To unsubscribe:    http://lists.mysql.com/maxdb?unsub=gcdm-maxdb@xxxxxxxxxxx