Having installed the the new Automated Predictive Library (APL), the former KXEN Infinite Insight Libraries inside HANA, I wanted to share my experience.
The benefit the APL brings is that we can now utilise the KXEN libraries directly in-memory within the HANA Platform so the data never needs to leave HANA. These libraries are a core part of Predictive Analytics 2.0 and are found in both the Automated and Expert perspectives within this. As the name suggests when using the APL we have attempted to automate as much of the predictive process as possible. To learn more about Predictive Analytics and the APL see
herve.kauffmann blog
Introducing the Automated Predictive Library (APL) for SAC AND HANA
With the Predictive Analytics 2.0 release we have actually released version 1.1 of the APL, (version 1.0 was an internal only release).
With the APL 1.1 the following predictive functions have been included.
- Classification
- Regression
- Clustering
- Time Series
- Key Influencers
For a deeper explanation of the APL please have a look at a blog post from
Ashish Morzaria
What is the SAP Automated Predictive Library (APL) for SAP HANA?
For the official SAP release information please see the
SAP Note 2111573.
The APL is actually contained within Predictive Analytics 2.0 download, so first you will need to
download that component first.
The official APL documentation can be found at
http://help.sap.com/pa
The pre-requisits for the APL are
1. SAP HANA 1.0 SPS 09 or above - I used HANA revision 92
2.
unixODBC - I downloaded and compiled from the source here. I won't repeat the instructions as they are very clear at
http://www.unixodbc.org/
3. AFL SDK 1.00.090 - You need don't need to install this as it already comes with HANA
Once unixODBC is installed you should find it in /usr/local/lib as shown below
As shown in the APL documentation on page7, you need to create a symbolic link for libodbc.so.1 such as
ln –s libodbc.so.2 libodbc.so.1
Now you need to add the /usr/local/lib to the LD_LIBRARY_PATH.
LD_LIBRARY_PATH=$LD_LIBRARY_PATH=:/usr/local/lib
These changes should be added to the customer.sh located in the home directory of the hana adm user. This file won't exist unless you have already made some changes to your environment.
The APL install itself is very easy, just upload and extract the TGZ file and then launch hdbinst as root. The file I uploaded was from the full PA 2.0 download and called "SAPPAAPL2000_0.TGZ"
mo-dad69e91a:/tmp/apl-1.1.0.20-linux_x64/installer # ls -l
total 28
-rwxr-xr-x 1 30003 shadow 19128 Nov 20 22:01 hdbinst
drwxr-xr-x 5 root root 4096 Feb 12 13:57 instruntime
drwxr-xr-x 2 root root 4096 Feb 12 13:57 packages
mo-dad69e91a:/tmp/apl-1.1.0.20-linux_x64/installer # ./hdbinst
Enter your SYSTEM ID and credentials and the APL will be installed, HANA will be restarted to allow the install to complete.
That is the software installed you now need to configure it by running the supplied SQL scripts.
If you navigate to the extracted TGZ location you should see samples/sql and in here you can open the apl_admin.sql as shown
In here there is some "sample" SQL code to use
🙂
You need to enable the HANA ScriptServer if it is not already running within your environment - either directly via SQL or in HANA Studio at Admin Console, Configuration, daemon.ini
alter system alter configuration ('daemon.ini', 'SYSTEM') set ('scriptserver', 'instances') = '1' with reconfigure;
With the scriptserver running you can now check the APL is installed correctly, all of the APL function statements should return APL related rows such as shown below
As you can see the APL is now installed, now you may want to install and configure the APL and the samples provided, which can be found here -
Using the APL samples provided with Predictive Analysis 2.0
If you are updating your HANA install then you may need to uninstall the APL, I've described that here
How to uninstall the Automated Predictive Library in SAP HANA