Database Replication
Posted in 1999
Topics: Performance & Tuning, Storage & Space Management
We want to use the Informix Enterprice Replication V7.3 in the asynchronous modus to do the following: The source database is located on a Windows NT server and is used for read/write access. This database has actually a size of 1 GB (250 MB dbspace and 750 MB blobspace). We want to replicate this data to 10 target databases for read only. All target databases are already exists on a Unix server. That means I want to replicate some tables from the source database on NT into several existing target databases on Unix. The changes on the source database are rarely. But if we make changes, we will insert a BLOB object about 15 MB. Can I get into troubles with the replication ? We have got a LAN-WAN-LAN network with 125 kb/sec baud rate. What is the best way to make a synchronisation in case of adding a new target database server and in case of losing data ? I want to use my network not a tape. Can I use the High Performance Loader to reload a 1 GB database ? Can anyone answer my questions ? With kind regards. I
With 125kb/sec baud, it's going to take quite a bit of time to transmit each of the blobs. 15Meg is 120Meg in bits. So that means that it's probably going to take about 15-20 minutes to transmit the blob. That's a long time in which network problems might occur. One of the things that you might want to consider is to get 7.31 in place rather than 7.30. There is a new network compression feature for ER in 7.31 which could come in handy here. Another thing to be aware of - blob columns are not replicated unless the blob column itself is updated. So if you are updating non-blob columns we don't encounter the overhead of transmitting the blobs. ---------------------------------------- As far as creating a new instance, yes - I'd probably use HPL. It's good that you are considering what to do with possible loss of data. It should never occur, but it's good that you are planning for the possibability. I'd be tempted to use HPL also in the event that there was loss of data. The question is, do I simply rebuild the instance, or do I selectivly try to recover. If it appears to be only a simple amount of data lost, I'd also consider doing a remote select from/insert into. Puntigam Franz wrote: > We want to use the Informix Enterprice Replication V7.3 in the > asynchronous modus to do the following: > The source database is located on a Windows NT server and is used for > read/write access. This database has actually a size of 1 GB (250 MB > dbspace and 750 MB blobspace). > We want to replicate this data to 10 target databases for read only. > All target databases are already exists on a Unix server. That means > I want to replicate some tables from the source database on NT into > several existing target databases on Unix. > > The changes on the source database are rarely. But if we make changes, > we will insert a BLOB object about 15 MB. > Can I get into troubles with the replication ? We have got a > LAN-WAN-LAN network with 125 kb/sec baud rate. > > What is the best way to make a synchronisation in case of adding a new > target database server and in case of losing data ? I want to use my > network not a tape. > Can I use the High Performance Loader to reload a 1 GB database ? > > Can anyone answer my questions ? > > With kind regards. > > I