I plan to use Enterprise Replication for replicating information amongst multiple IDS instances. Regarding this I had few queries:
Can there be two different sets of replicates configured between different nodes. For e.g. consider the below scenario:
Replicate 1 - Nodes A and B replicate table X information amongst themselves
Replicate 2 - Nodes A, B and C replicate table Y information amongst themselves
Whats the level of scalability offered by ER in terms of the number of nodes which can be inter-connected in Update - Anywhere topology. Also, whats the scalability offered in terms of number of bytes which can be transferred.
Is ER IPv6 compliant. I mean the interface on which replication takes place, can that be IPv6 interface.
What are the licensing terms for ER. Do I need a separate license for ER if I already have license for normal IDS
How costly is it to replicate blobs.
Is ER FIPS compliant
Last edited by gssachdeva; 01-18-12 at 08:40.
Reason: Further queries
as a response to some of your questions:
* IPV6: yes this is supported
* licensing terms for ER: on a generic point of view, this depends on the version of IDS and edition you have. For all 11.70 editions, you have ER included. The smallest edition (Innovator-C, the free one), gives you access to two ER nodes. Starting from the Growth Edition, ER is unlimited. I recommand this very good comparative document written by Carlton Doe to understand how the thing is.
For the former Editions, having your current version number would help to identify better what you are entitled to.
* scalability: this depends on the IDS Edition you have and the resource of each of your servers, namedl processors/cores number and available memoryy. As you know, Informix Enterprise Replication is based on a logical logs capture and evaluation, Network transport and finally apply on the taget server(s). The logical logs reading and evaluation are handled in memory by the Reliable Queue Manager (a specialized IDS thread), which can overload if logical logs space is too small or memory allocated too small. In any case, if overloading, data is written to a preconfigured dbspace in order not to be lost. The mechanism is the same for the receive queue.
A weak network can be a contention point.
As long as network is fast and the RQM has enough memory, your ER will have a good thruput rate. But again, the evaluation phase will optimize the amount of data to be transported. For instance, if within the same transaction a row is inserted, then updated, only the final updated row image will be transported, limiting the number of bytes transported on the network.
In fact a really briliant implementation of replication!
Regarding blobs, you will have to assign one or more sbspaces to spool then ( they would blow out your RQM memory ). Obviously this will generate some IO, as well as more network traffic, but in my sense this is the fastest way to achieve this functionality safely.
* As per Release Notes, Informix ER seems to be FIPS compliant since 11.50xC6
* Regarding the combinations, I would say that the combinations you mention are possible, playing with the replication servers and the cdr define command, because a replication schema is based on a table/columns basis.