RE: Enterprise Replication / HDR (2855)
Posted in 2000
This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01BF6919.7085BD18 Content-Type: text/plain; charset="iso-8859-1" Keep in mind with ER -- you may not get the "workload split" you are hoping for if everything gets replicated immediately, because anything that is changed on server A must ALSO be changed (via replication) on server B, and vice versa. Plus you will now have the extra overhead on each server of ER itself. You may be able to alleviate this by scheduling the "synchronization" of the servers for off-peak -- but then you take the *BIG* gamble of data getting out of sync between the two servers. What ER *CAN* help with (especially in update-anywhere mode) is avoiding a single-point-of-failure (SPOF). If server A goes down, simply switch all users over to server B. Ideally, this switching process can be automated. We have looked at ER quite a lot, and have it implemented on several 7.3x systems, primarily for this SPOF reason. If you've got one group of users that do most of the updating, and another group that simply run queries / reports, then HDR might be a good choice. You could direct the one set of users to server A, and the other set to server B. And HDR itself has pretty low overhead. We have it running on a very mission critical application, and have often had to do a switchover (take all users off of primary, take primary down, change secondary to standalone mode, and move all users to the new box), and this whole process, even done manually, takes only a couple of minutes. Can't say anything about Sblobs, though, as we're still on 7.2x. :-) HTH, Paul Mosser -----Original Message----- From: James Maes [mailto:jmaes@sportingnews.com] Sent: Thursday, January 27, 2000 2:49 PM To: informix-list@iiug.org Subject: Enterprise Replication / HDR (2855) Hey all, here is one for you all that test support has as of so far been unable to help with. I am planning / setting up a new database of the following Informix DS 9.20UC1 on 2 linux 2.2.5-15 machines (everything the same) The database is used in a publishing system, and will have the Web Datablades installed. The configuration that I am looking into is one of the following, mind you, the goal of this is to spilt the work load between the two mahines Configuration A haveing the machines in Enterprise Replication useing Update Anywhere model, the only problem is that ER does not replicate Smart Blobs the web data blades use smart blobs so this model will not work unless I can find a get smart blobs to relicate to both machines Configuration B having the machines doing HD Replication this forces the primary machine into read/write mode and the secondary machine in Read Only mode. this will replicate the data from the two machines okay, but makes some other things harder to do. I would love any feed back from anyone out there that has done something like this. Thanks -- James Maes unix/database programmer jmaes@sportingnews.com The Sporting News www.sportingnews.com ------_=_NextPart_001_01BF6919.7085BD18 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> <HTML> <HEAD> <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; = charset=3Diso-8859-1"> <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = 5.5.2448.0"> <TITLE>RE: Enterprise Replication / HDR (2855)</TITLE> </HEAD> <BODY> <P><FONT SIZE=3D2>Keep in mind with ER -- you may not get the = "workload split" you are hoping for if everything gets = replicated immediately, because anything that is changed on server A = must ALSO be changed (via replication) on server B, and vice = versa. Plus you will now have the extra overhead on each server = of ER itself. You may be able to alleviate this by scheduling the = "synchronization" of the servers for off-peak -- but then you = take the *BIG* gamble of data getting out of sync between the two = servers. What ER *CAN* help with (especially in update-anywhere = mode) is avoiding a single-point-of-failure (SPOF). If server A = goes down, simply switch all users over to server B. Ideally, = this switching process can be automated. We have looked at ER = quite a lot, and have it implemented on several 7.3x systems, primarily = for this SPOF reason.</FONT></P> <P><FONT SIZE=3D2>If you've got one group of users that do most of the = updating, and another group that simply run queries / reports, then HDR = might be a good choice. You could direct the one set of users to = server A, and the other set to server B. And HDR itself has = pretty low overhead. We have it running on a very mission = critical application, and have often had to do a switchover (take all = users off of primary, take primary down, change secondary to standalone = mode, and move all users to the new box), and this whole process, even = done manually, takes only a couple of minutes.</FONT></P> <P><FONT SIZE=3D2>Can't say anything about Sblobs, though, as we're = still on 7.2x. :-)</FONT> </P> <P><FONT SIZE=3D2>HTH,</FONT> <BR><FONT SIZE=3D2>Paul Mosser</FONT> </P> <P><FONT SIZE=3D2>-----Original Message-----</FONT> <BR><FONT SIZE=3D2>From: James Maes [<A = HREF=3D"mailto:jmaes@sportingnews.com">mailto:jmaes@sportingnews.com</A>= ]</FONT> <BR><FONT SIZE=3D2>Sent: Thursday, January 27, 2000 2:49 PM</FONT> <BR><FONT SIZE=3D2>To: informix-list@iiug.org</FONT> <BR><FONT SIZE=3D2>Subject: Enterprise Replication / HDR (2855)</FONT> </P> <BR> <P><FONT SIZE=3D2>Hey all, here is one for you all that test support = has as of so far been</FONT> <BR><FONT SIZE=3D2>unable to help with.</FONT> </P> <BR> <P><FONT SIZE=3D2>I am planning / setting up a new database of the = following</FONT> </P> <P><FONT SIZE=3D2>Informix DS 9.20UC1 on 2 linux 2.2.5-15 machines = (everything the same)</FONT> </P> <P><FONT SIZE=3D2>The database is used in a publishing system, and will = have the Web</FONT> <BR><FONT SIZE=3D2>Datablades installed.</FONT> </P> <P><FONT SIZE=3D2>The configuration that I am looking into is one of = the following,</FONT> </P> <P><FONT SIZE=3D2>mind you, the goal of this is to spilt the work load = between the two</FONT> <BR><FONT SIZE=3D2>mahines</FONT> </P> <P><FONT SIZE=3D2>Configuration A</FONT> </P> <P><FONT SIZE=3D2>haveing the machines in Enterprise Replication = </FONT> <BR> <FONT SIZE=3D2>useing = Update Anywhere model, the only problem is that ER does not</FONT> <BR><FONT SIZE=3D2>replicate Smart Blobs</FONT> <BR> <FONT SIZE=3D2>the web = data blades use smart blobs</FONT> <BR> &