IWETHEY v. 0.3.0 | TODO
1,095 registered users | 0 active users | 0 LpH | Statistics
Login | Create New User
IWETHEY Banner

Welcome to IWETHEY!

New Could you do multi-master replication for < $1-million?
We currently serve our intranet from a single web server connected to about a dozen DB servers. There are several primaries, each with one or more replicated slaves. As we are approaching saturation of the T1s into our remote offices, it's become clear we need to move to a distributed system. We plan to put web servers in each branch office, served by local DB servers. The problem is replication.

We already have pages that have to read from the master: since the one-way replication in MySQL doesn't allow verification that the change has been propogated to each of the slaves, we have to read from the master so users see their changes in real time. And this is with all servers in the same rack on Gig ethernet. The problem would be orders of magnitude worse when the slaves aren't even in the same state as the master.

We are going to be soliciting resumes for a DBA.[1] The interview will consist of describing our surrent setup and future needs, and ask what they would do and why. Whoever makes the best case for their solution gets hired to implement it. Could be MaxDB[2], Oracle[3], DB2[4], PostgreSQL[5], anything that will work[6].

Oh, and the $1-million in the subject line? That's what it would cost to install Oracle to do what we want.


[1] I'm not making the hiring decision, but can forward resumes.

[2] Formerly SABDB, now distributed by MySQL-AG. This has the advantages that we're already on MySQL, and that it's GPL.

[3] The no-brainer solution that we know would work by throwing money at it.

[4] Scales from blade servers to i390s.

[5] White papers from 2 years back indicated multi-master was close on Postgres, but I can't find anything saying it's working.

[6] Conspicuously absent is SQL Server. Virus-caused downtime is a killer.
===

Implicitly condoning stupidity since 2001.
New Sure, T3's :-)
"You're just like me streak. You never left the free-fire zone.You think aspirins and meetings and cold showers are going to clean out your head. What you want is God's permission to paint the trees with the bad guys. That wont happen big mon." Clete
questions, help? [link|mailto:pappas@catholic.org|email pappas at catholic.org]
New /me smacks forehead
Well no shit. Yeah, that would push the problem back a couple of months at least. However, we still want the branch offices to be able to keep functioning even if other offices go offline, like the blackout last year. (I know, I know ... I didn't include that in the initial spec.)
===

Implicitly condoning stupidity since 2001.
New Well from a support guy point of view
distributed Webservers pointed at centralized DB's would be easier to manage. 2 geographically dispersed hosted in ironman datacenters, one master and one slave with enough bandwidth. Webservers in each decentralized location with a caching solution like weblogic so the clients are thin. In a blackout your ohio DB would autofail to the texas database. Cheaper than Oracle.
thanx,
bill
"You're just like me streak. You never left the free-fire zone.You think aspirins and meetings and cold showers are going to clean out your head. What you want is God's permission to paint the trees with the bad guys. That wont happen big mon." Clete
questions, help? [link|mailto:pappas@catholic.org|email pappas at catholic.org]
New That's one of the things we're considering
Like I said, we'll consider anything. But looking at the cost of T3s ... wooh.
===

Implicitly condoning stupidity since 2001.
New well with the thin clients pointing to an webserver at the
colo, your traffic would not be as heavy as an app server at the remote location. Your appserver(s) would be at the colo with only screen updates being ferried to the remotes. Might not need T3's
thanx,
bill
"You're just like me streak. You never left the free-fire zone.You think aspirins and meetings and cold showers are going to clean out your head. What you want is God's permission to paint the trees with the bad guys. That wont happen big mon." Clete
questions, help? [link|mailto:pappas@catholic.org|email pappas at catholic.org]
New Does this answer your question about PG?
[link|http://developer.postgresql.org/todo.php|PG TODO]

More work being devoted to this for you right now by me.

--
[link|mailto:greg@gregfolkert.net|greg],
[link|http://www.iwethey.org/ed_curry|REMEMBER ED CURRY!] @ iwethey

I did a 10K wheelchair race once. The NTMFAC
who pushed me still has the whip-marks.
New Much cool
Any idea on when it would be available?
===

Implicitly condoning stupidity since 2001.
New Willingness to spend some of that <$1million might help :-P
"good ideas and bad code build communities, the other three combinations do not"
- [link|http://archives.real-time.com/pipermail/cocoon-devel/2000-October/003023.html|Stefano Mazzocchi]
New We are ... ___________________________________ No, seriously
If someone comes in and convinces us he can finish replication in Postgres for less than whatever it would cost for another solution, we'd consider it. Bearing in mind our comfort level with a promise vs. a known cost for the alternatives. Yes, self-FUD, but entirely real uncertainty, not manufactured.
===

Implicitly condoning stupidity since 2001.
New Goto:
[link|http://advocacy.postgresql.org|http://advocacy.postgresql.org]

And do what you need to do to make them know you are seriously looking at spending money.

Of course the greatest thing would be to GNU GPLv2+ it as well.

Or you could also subscribe to some [link|http://www.postgresql.org/lists.html|Mailing Lists]

Or Heck, even pay for a contract to DO the work and guarantee updates to it and so on. All ya gotta do is ask. PG will scale and it feels more righter than MySQL does a lot of the time.
--
[link|mailto:greg@gregfolkert.net|greg],
[link|http://www.iwethey.org/ed_curry|REMEMBER ED CURRY!] @ iwethey

I did a 10K wheelchair race once. The NTMFAC
who pushed me still has the whip-marks.
New Will do, thanks for the pointers
===

Implicitly condoning stupidity since 2001.
     Could you do multi-master replication for < $1-million? - (drewk) - (11)
         Sure, T3's :-) -NT - (boxley) - (1)
             /me smacks forehead - (drewk)
         Well from a support guy point of view - (boxley) - (2)
             That's one of the things we're considering - (drewk) - (1)
                 well with the thin clients pointing to an webserver at the - (boxley)
         Does this answer your question about PG? - (folkert) - (5)
             Much cool - (drewk) - (4)
                 Willingness to spend some of that <$1million might help :-P -NT - (ben_tilly) - (1)
                     We are ... ___________________________________ No, seriously - (drewk)
                 Goto: - (folkert) - (1)
                     Will do, thanks for the pointers -NT - (drewk)

Mmmmmmm.... Nemo!
106 ms