October 18, 2012
Qlikview can go outside RAM, finally
Posted by Andrei Pandre under DV News
| Tags: Qlikview
Leave a Comment
Qlikview 10 was released around 10/10/10, Qlikview 11 – around 11/11/11, so I expected Qlikview 12 to be released on 12/12/12 but “instead” we are getting Qlikview 11.2 with Direct Discovery in December 2012, which supposedly provides a “hybrid approach so business users can get the QlikView associative experience even with data that is not stored in memory”
This feature demanded by users (me included) for a long time, but I think noise around so called Big Data and competition forced Qliktech to do it. Spotfire has it for a longtime (as well as 64-bit implementation) and Tableau has something like that for a while (unfortunately Tableau still 32-bit) . You can test Beta of it, if you have time: http://community.qlikview.com/blogs/technicalbulletin/2012/10/22/qlikview-direct-discovery-beta-registration-is-open
Just 8 months ago Qliktech estimated its sales for 2012 as $410M and suddenly 3 months ago it changed its estimates down to $381M, just 19% over 2011, which is in huge contrast with Qliktech’s previous speed of growth and way behind the current speed of growth of Tableau and even less then current speed of growth of Spotfire. During last 2 years QLIK stock unable to grow significantly:
and all of the above forcing Qliktech to do something outside of gradual improvements – new and exciting functionality needed and Direct Discovery may help!
QlikView Direct Discovery enables users to perform visual analysis against “any amount of data, regardless of size”. With the introduction of this unique hybrid approach, users can associate data stored within big data sources directly alongside additional data sources stored within the QlikView in-memory model. QlikView can “seamlessly connect to multiple data sources together within the same interface”, e.g. Teradata to SAP to Facebook allowing the business user to associate data across the data silos. Data outside of RAM can be joined with the in-memory data with the common field names. This allows the user associatively navigate both on the direct discovery and in memory data sets.
QlikView developer should setup the Direct Discovery table on the QlikView application load script to allow the business users to query the desired big data source. Within the script editor a new syntax is introduced to connect to data in direct discovery form. Traditionally the following syntax is required to load data from a database table:
To invoke the direct discovery method, the keyword “SQL” is replaced with “DIRECT”.
In the example above only column CarrierTrackingNumber and ProductID are loaded into QlikView in the traditional manner, other columns exist in the data table within the Database including columns OrderQty and Price. OrderQty and Price fields are referred as “IMPLICIT” fields. An implicit field is a field that QlikView is aware of on a “meta level”. The actual data of an implicit field resides only in the database but the field may be used in QlikView expressions. Looking at the table view and data model of the direct discovery columns are not within the model (on the OrderFact table):
Once the direct discovery structure is established, the direct discovery data can be joined with the in-memory data with the common field names (Figure 3). In this example, “ProductDescription” table is loaded in-memory and joined to direct discovery data with the ProductID field. This allows the user to associatively navigate both on the “direct discovery” and in memory data sets.
Direct Discovery will be much slow then in-memory processing and this is is expected, but it will take away from Qlikview its usual claim that is is faster then competitors. QlikView Direct Discovery can only be used against SQL compliant data sources. The following data sources are supported;
• ODBC/OLEDB data sources – All ODBC/OLEDB sources are supported, including SQL Server, Teradata and Oracle.
• Custom connectors which support SQL – Salesforce.com, SAP SQL Connector, Custom QVX connectors for SQL compliant data stores.
Due to the interactive and SQL syntax specific nature of the Direct Discovery approaches a number of limitations exist. The following chart types are not supported;
• Pivot tables
• Mini charts
And the following QlikView features are not supported:
• Advanced aggregation
• Calculated dimensions
• Comparative Analysis (Alternate State) on the QlikView objects that use Direct
• Direct Discovery fields are not supported on Global Search
• Binary load from a QlikView application with Direct Discovery table
Here is a some preliminary video about Direct Discovery, published by Qliktech:
It was interesting to me that just 2 days after Qliktech pre-anounced Direct Discovery it also partners with Teradata. Tableau partners with Teradata for a while and Spotfire did it a month ago, so I guess Qliktech trying to catchup in this regard as well. I mentioned it only to underscore the point of this blog post: Qliktech realized that it behind its competitors in some areas and it has to follow ASAP.