Thread: Need help for our thesis.

As luck would have I've put some thought into this problem before as 
well.  Having reviewed the code quite a bit, and played around with the 
backend, my intuition tells me that the following would work.

Write a custom "cluster-aware" postmaster so it distributes queries to 
various machines.  Of course each machine would have to be running a 
normal postmaster to receive the queries.  And, the cluster-aware 
postmaster would need to be informed of which machines to submit 
queries to (e.g. listed in a configuration file).  Queries could be 
processed in the following ways:
* database management queries (e.g. create database, create 
table, alter table, etc. . . ) would be distributed to every machine* select queries would be distributed to every
machine. The 
 
cluster-aware postmaster would merge the results before delivering them 
to the client.* an insert query would go to one machine only, and the 
cluster-aware postmaster could switch between the designated machines 
after each insert.  For example, given two machines, the first insert 
would go to machine A, the second to machine B, etc . . .

Don't know if that's been tried, but that's where I would start.  
Postgres' modular design of a postmaster and a backend would make this 
relatively easy to do, because the only thing you might end up writing 
is the one cluster-aware postmaster.  If you end up doing this let me 
know and if I have time I'll help out with the design/coding.

Richard Schilling
Lake Stevens, WA USA



On 2003.07.04 18:56 MIka Santos wrote:
> Gud day! We are currently having our thesis for our undergarduate
> course and this involoves the code of PostgreSQL. Basically, our
> thesis needs a modification of the existing code of the said database.
> As of now, our University has an existing online registration system
> which uses Postgre as their back end. But the problem is the existing
> one is not a distributed system. Our thesis adviser would like us to
> deal with the load balancing of the said online registration. He
> wanted a multidatabase querying. For example, we knew that the
> following statements are posible,
>                    select *
>                    from t1, t2., t3;
> provided that t1 and t2, t3 are tables from a single database.
> Our thesis is to make the following satatement legal
>                   select *
>                   from d1, d2, d3;
> provided that d1 and d2, d3 are separate databases.
> We are currently having our research about this problem and as
> beginners, we surely need help from experts like you. Any form of
> comments and suggestions will surely be appreciated by the group.
>       Thank you very much and hope to here from you.
> 
> 
> ---------------------------------
> Do you Yahoo!?
> Free online calendar with sync to Outlook(TM).

Hello Mika

Currently you can use dblink which queries a remote database and returns 
the data as a table. This works nicely for PostgreSQL. There is also a 
JDBC version in progress.
We have writte an Oracle version of dblink.

SELECT * FROM dblink('connectstr', 'SELECT ...') AS some_type;

This works nicely.
Still, this is not a full implementation of distributed queries - this 
would be a huge project. There has already been some discussion about 
that on the hackers list.
Regards,
    Hans



Richard Schilling wrote:
> As luck would have I've put some thought into this problem before as 
> well.  Having reviewed the code quite a bit, and played around with the 
> backend, my intuition tells me that the following would work.
> 
> Write a custom "cluster-aware" postmaster so it distributes queries to 
> various machines.  Of course each machine would have to be running a 
> normal postmaster to receive the queries.  And, the cluster-aware 
> postmaster would need to be informed of which machines to submit queries 
> to (e.g. listed in a configuration file).  Queries could be processed in 
> the following ways:
> 
>     * database management queries (e.g. create database, create table, 
> alter table, etc. . . ) would be distributed to every machine
>     * select queries would be distributed to every machine.  The 
> cluster-aware postmaster would merge the results before delivering them 
> to the client.
>     * an insert query would go to one machine only, and the 
> cluster-aware postmaster could switch between the designated machines 
> after each insert.  For example, given two machines, the first insert 
> would go to machine A, the second to machine B, etc . . .
> 
> Don't know if that's been tried, but that's where I would start.  
> Postgres' modular design of a postmaster and a backend would make this 
> relatively easy to do, because the only thing you might end up writing 
> is the one cluster-aware postmaster.  If you end up doing this let me 
> know and if I have time I'll help out with the design/coding.
> 
> Richard Schilling
> Lake Stevens, WA USA
> 
> 
> 
> On 2003.07.04 18:56 MIka Santos wrote:
> 
>> Gud day! We are currently having our thesis for our undergarduate
>> course and this involoves the code of PostgreSQL. Basically, our
>> thesis needs a modification of the existing code of the said database.
>> As of now, our University has an existing online registration system
>> which uses Postgre as their back end. But the problem is the existing
>> one is not a distributed system. Our thesis adviser would like us to
>> deal with the load balancing of the said online registration. He
>> wanted a multidatabase querying. For example, we knew that the
>> following statements are posible,
>>                    select *
>>                    from t1, t2., t3;
>> provided that t1 and t2, t3 are tables from a single database.
>> Our thesis is to make the following satatement legal
>>                   select *
>>                   from d1, d2, d3;
>> provided that d1 and d2, d3 are separate databases.
>> We are currently having our research about this problem and as
>> beginners, we surely need help from experts like you. Any form of
>> comments and suggestions will surely be appreciated by the group.
>>       Thank you very much and hope to here from you.


-- 
Cybertec Geschwinde u Schoenig
Ludo-Hartmannplatz 1/14, A-1160 Vienna, Austria
Tel: +43/2952/30706; +43/664/233 90 75
www.cybertec.at, www.postgresql.at, kernel.cybertec.at

On 4 Jul 2003 at 18:56, MIka Santos wrote:
> said online registration. He wanted a multidatabase querying. For example, we 
> knew that the following statements are posible,
>  select *
>  from t1, t2., t3;
> provided that t1 and t2, t3 are tables from a single database.
> Ourthesis is to make the following satatement legal
>  select *
>  from d1, d2, d3;
> provided that d1 and d2, d3 are separate databases.

Would schema help you in anyways?

ByeShridhar

--
Van Roy's Law:    An unbreakable toy is useful for breaking other toys.