[Re-posted due to blog issues]
Informix 11.70.FC5 has three of significant features for Informix warehouse accelerator (IWA).
1. Supporting Informix host running on Solaris/Intel. IWA will still run on Linux/Intel on a separate hardware.
2. Partition based refresh from Informix to IWA
3. Deploying and using IWA in various HA cluster configurations (Primary, SDS, HDR secondary, RSS, etc).
The few slides below give you a straight forward explanation of the two main use cases.
fyi: You can download the power point version with animation at: http://slidesha.re/KnfTYK
I've added commentary for each slide below.
is the IWA architecture and steps to deploy and use IWA till
11.70.FC4.. IWA takes a snapshot of the data from Informix and runs
queries on that data. The transactions and loads happening on Informix
won't immediately change the data on IWA. To do that, you've to refresh
the data either using the studio or command line tool. When you
refresh in this method, you'll reload the entire data. Time to
completely refresh is directly correlated to size of your data set.
Informix 11.70.FC5 alleviates this issue by allowing you to ONLY refresh the changed partitions of the fact table.
Slide 2: The
Sales table is the fact table in this mart. After you've setup the
datamart in IWA, your load jobs will load data into one of the
partitions or add new paritions into the fact table. In this case, you
simply do the following:
1. execute dropPartMart() procedure for each of previously existing partitions that were modified.
2. execute loadPartMart() procedure for each of modified and new partitions.
do have to know or keep track of which of the partitions in your fact
table were modified since the last refresh and which of the new
partitions are added.
Time cyclic data management enables you to limit the amount of data you
keep in your data marts. If your business requires a 3 week window, at
the start of every week, you roll off (detach) the oldest week and
attach the data from the latest week.
This slide is better seen with animation in power point. In step 2,
execute the dropPartMart() to drop the IWA data for the partition you're
about to detach and then detach the partition. After you attach the
new partition, issue the loadPartMart() to refresh the data to IWA.
When you have large data marts, using the partition based refresh
will enable you to refresh the data quickly and resume your analysis.
Because you can refresh the data quickly, you can choose to refresh the
data often to enable analysis with latest data.
This is a step towards eventual goal of near/real time analysis.