Re: Transparent replication of DB2 data on WAN ?
Date: Mon, 16 Aug 1993 22:18:59 GMT
Message-ID: <CBvHzn.3M6_at_csfb1.fir.fbc.com>
In article <24o60n$6k9_at_DIALix.oz.au>, manfan_at_DIALix.oz.au (Manuel Tsz Pong Fan) writes:
|>
|> Our organization is in the process of moving away from an architecture
|> which relies solely on DB2/MVS to store and CICS/MVS to access trusted
|> corporate data. The option currently favoured by management and heavy
|> weight users is to replicate DB2 data onto DOZENS of databases on a WAN.
|>
|> Before I double my life insurance, can anyone tell me whether any
|> vendor provides a solution which will more or less transparently
|> synchronize replicas either continuously or at pre-determined intervals?
|> Has anyone actually done this?
|>
|> Can System 10 do it?
|>
|> Can Oracle handle replication transparently at all?
|>
|> BTW, is any DBMS vendor besides IBM supporting or planning to support
|> DRDA?
|>
|>
|> manuel fan
|> programmer analyst
|> water authority of western australia
|> voice: 61-9-4202014
|> fax: 61-9-4203177
|> internet: manfan_at_DIALix.oz.au
With Sybase's replication server there is a 'log transfer manager' that reads Sybase transaction logs and feeds the update information to the replication mechenism. You can feed it from other sources, but there is no log_transfer manager for DB2. This means that you would have to feed the replicator your DB2 updates through some mechenism of your own creation. This might be done by reading the DB2 recovery logs ( does IBM documnt them well enough for this??) or capture the database updates before they get to DB2.
With Oracle, (speak up if you know differently folks) I believe they deal with periodic replication of table snapshots. Hopefully there is a mechenism for frequent incremental copies. It does not appear to work through a transaction log mechenism. This means that you would once again have to capture your DB2 updates and feed them into an Oracle database before the Oracle replication could work.
Informix has a DRDA gateway with RUW (Remote Unit of Work) just entering production now. It can be used with several of the DB2 products with DB2/MVS being the primary target. It cannot do distributed joins that include DB2 tables and Informix tables since they require an Informix catalog to do their distributed operations. I suspect that some of their 4GL type products would have similar problems.
Our company currently copies some complete databases overnight, uses timestamps to do incremental extracts of others and uses a replication mechenism for real-time replication of others. (homegrown replicator) Most databases are replicated to two Unix servers with a few being replicated to as many as four servers. Most of the Unix replias are table subsets of the mainframe data with tables being set up on Unix on as needed basis. We have a few tables which are subsetted as well, both vertically and horizontally.
I would look closely at the network bandwidth that will be needed for replicating your databases. Most likely, the current network is insufficient to support current loads as well as replication. This coupled with a cost analysis (look at the costs of the Sybase software as well) should serve to reduce the number of replicas that they plan for and would buy you some time for evaluation as well.
Good Luck
" " " Lawrence
Lawrence E. Hall
First Boston Corp.
5 World Trade Center 9th Floor uunet!csfb1!phantom!lhall NYC, NY 10048 lhall_at_csfb1.fir.fbc.com ----------------------------------------------------------------------------This message will self-destruct in 5 seconds. Received on Tue Aug 17 1993 - 00:18:59 CEST