public inbox for [email protected]  
help / color / mirror / Atom feed
Distributed Database Web Site
5+ messages / 4 participants
[nested] [flat]

* Distributed Database Web Site
@ 2004-08-14 20:50 Marc G. Fournier <[email protected]>
  2004-08-15 10:23 ` Re: Distributed Database Web Site Justin Clift <[email protected]>
  2004-08-16 14:26 ` Re: Distributed Database Web Site Alexey Borzov <[email protected]>
  0 siblings, 2 replies; 5+ messages in thread

From: Marc G. Fournier @ 2004-08-14 20:50 UTC (permalink / raw)
  To: pgsql-www


Just curious, how hard would it be to distinguish on the main web site 
between read and write operations?

For instance, if we had a read-only database on a mirror site, replicated 
from the main database using something like Slony ... how hard would it be 
t have 'write operations' sent to the main web site to happen, with all 
read operations being local?

ie. posting news events would happen at http://www.postgresql.org, but 
reading news events would be @ http://www.xx.postgresql.org, but both 
could/would be dynamic?

----
Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
Email: [email protected]           Yahoo!: yscrappy              ICQ: 7615664



^ permalink  raw  reply  [nested|flat] 5+ messages in thread

* Re: Distributed Database Web Site
  2004-08-14 20:50 Distributed Database Web Site Marc G. Fournier <[email protected]>
@ 2004-08-15 10:23 ` Justin Clift <[email protected]>
  2004-08-15 15:50   ` Re: Distributed Database Web Site Marc G. Fournier <[email protected]>
  1 sibling, 1 reply; 5+ messages in thread

From: Justin Clift @ 2004-08-15 10:23 UTC (permalink / raw)
  To: Marc G. Fournier <[email protected]>; +Cc: pgsql-www

Marc G. Fournier wrote:
 >
> Just curious, how hard would it be to distinguish on the main web site 
> between read and write operations?
> 
> For instance, if we had a read-only database on a mirror site, 
> replicated from the main database using something like Slony ... how 
> hard would it be t have 'write operations' sent to the main web site to 
> happen, with all read operations being local?

Thinking about what it would take...

We'd need the pg_hba.conf of the main database set to allow connections 
from the mirror website(s), so the write operations can occur.

Or, perhaps as an alternative, we could look at making the pages that 
generate writes only be on the main PG website.  So, the majority of 
content is mirrored except for a few pages which do stuff like (i.e.) 
submit new stories and stuff.

I haven't really played with Slony in the scenario of multiple mirrors 
that may or may not all be available at any one time, but Jan should be 
able to give input here.


> ie. posting news events would happen at http://www.postgresql.org, but 
> reading news events would be @ http://www.xx.postgresql.org, but both 
> could/would be dynamic?

Yep, should be possible.  We just have to think about the technology 
we'd like web mirrors to run, and then ensure it's acceptable and do-able.

:)

Regards and best wishes,

Justin Clift


> ----
> Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
> Email: [email protected]           Yahoo!: yscrappy              ICQ: 7615664
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
>    (send "unregister YourEmailAddressHere" to [email protected])




^ permalink  raw  reply  [nested|flat] 5+ messages in thread

* Re: Distributed Database Web Site
  2004-08-14 20:50 Distributed Database Web Site Marc G. Fournier <[email protected]>
  2004-08-15 10:23 ` Re: Distributed Database Web Site Justin Clift <[email protected]>
@ 2004-08-15 15:50   ` Marc G. Fournier <[email protected]>
  0 siblings, 0 replies; 5+ messages in thread

From: Marc G. Fournier @ 2004-08-15 15:50 UTC (permalink / raw)
  To: Justin Clift <[email protected]>; +Cc: Marc G. Fournier <[email protected]>; pgsql-www

On Sun, 15 Aug 2004, Justin Clift wrote:

> Marc G. Fournier wrote:
>>
>> Just curious, how hard would it be to distinguish on the main web site 
>> between read and write operations?
>> 
>> For instance, if we had a read-only database on a mirror site, replicated 
>> from the main database using something like Slony ... how hard would it be 
>> t have 'write operations' sent to the main web site to happen, with all 
>> read operations being local?
>
> Thinking about what it would take...
>
> We'd need the pg_hba.conf of the main database set to allow connections from 
> the mirror website(s), so the write operations can occur.
>
> Or, perhaps as an alternative, we could look at making the pages that 
> generate writes only be on the main PG website.  So, the majority of content 
> is mirrored except for a few pages which do stuff like (i.e.) submit new 
> stories and stuff.

This second was what I was thinking ...

> Yep, should be possible.  We just have to think about the technology 
> we'd like web mirrors to run, and then ensure it's acceptable and 
> do-able.

Now, knowing that all mirrors couldn't do this (or wouldn't), I was 
thinking of having a pseudo-regional master mirror that others could rsync 
static content from ...

----
Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
Email: [email protected]           Yahoo!: yscrappy              ICQ: 7615664



^ permalink  raw  reply  [nested|flat] 5+ messages in thread

* Re: Distributed Database Web Site
  2004-08-14 20:50 Distributed Database Web Site Marc G. Fournier <[email protected]>
@ 2004-08-16 14:26 ` Alexey Borzov <[email protected]>
  1 sibling, 0 replies; 5+ messages in thread

From: Alexey Borzov @ 2004-08-16 14:26 UTC (permalink / raw)
  To: Marc G. Fournier <[email protected]>; +Cc: pgsql-www

Hi,

Marc G. Fournier wrote:
> For instance, if we had a read-only database on a mirror site, 
> replicated from the main database using something like Slony ... how 
> hard would it be t have 'write operations' sent to the main web site to 
> happen, with all read operations being local?

What's the point? You can't realistically expect permanent high-speed 
connectivity between master and slaves, thus replication will be *extremely* 
asynchronous.

The only obviuos benefit I see is reduced traffic.

> ie. posting news events would happen at http://www.postgresql.org, but 
> reading news events would be @ http://www.xx.postgresql.org, but both 
> could/would be dynamic?

It works this way even now.




^ permalink  raw  reply  [nested|flat] 5+ messages in thread

* Re: Distributed Database Web Site
@ 2004-08-16 08:27 Dave Page <[email protected]>
  0 siblings, 0 replies; 5+ messages in thread

From: Dave Page @ 2004-08-16 08:27 UTC (permalink / raw)
  To: Marc G. Fournier <[email protected]>; Justin Clift <[email protected]>; +Cc: pgsql-www




-----Original Message-----
From: [email protected] on behalf of Marc G. Fournier
Sent: Sun 8/15/2004 4:50 PM
To: Justin Clift
Cc: Marc G. Fournier; [email protected]
Subject: Re: [pgsql-www] Distributed Database Web Site
 
> Now, knowing that all mirrors couldn't do this (or wouldn't), I was 
> thinking of having a pseudo-regional master mirror that others could rsync 
> static content from ...

What's the point? The main site uses a trivial amount of database access anyway, so unless you wanted to hack gforge for replication/static content mirroring...

Regards, Dave




^ permalink  raw  reply  [nested|flat] 5+ messages in thread


end of thread, other threads:[~2004-08-16 14:26 UTC | newest]

Thread overview: 5+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2004-08-14 20:50 Distributed Database Web Site Marc G. Fournier <[email protected]>
2004-08-15 10:23 ` Justin Clift <[email protected]>
2004-08-15 15:50   ` Marc G. Fournier <[email protected]>
2004-08-16 14:26 ` Alexey Borzov <[email protected]>
2004-08-16 08:27 Re: Distributed Database Web Site Dave Page <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox