Like most applications, the place where an analytics application gets developed and eventually deployed are rarely one and the same. An application may be developed in the cloud and then deployed locally or vice versa. The problem is that all those applications invariably need to be refactored when moving from one IT environment to another. IBM today moved to eliminate that issue with the launch of an Integrated Analytics System based on an instance of the Apache Spark in-memory computing framework that can migrate in and out of an IBM Cloud at the touch of a button.
Rob Thomas, general manager for IBM Analytics, says the Integrated Analytics System is an extension of the IBM Data Science Experience, which combines an instance of Apache Spark with a data warehouse based on a IBM DB2 relational database to enable analytics to be processed in real time alongside transactions.
The Integrated Analytics System should enable IT organizations to more efficiently take advantage of a hybrid cloud computing environment using a common base of software. Thomas says many IT organizations are unable to achieve that goal using service on, for example, Amazon Web Services (AWS) that have no on-premises equivalent.
“We think that approach is fundamentally broken,” says Thomas.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=i
In addition to making it simpler to migrate analytics applications, Thomas says, IBM’s major differentiation in this space is the cognitive services it makes available to analytics. There is no shortage of algorithms that can be employed to drive an analytics application. IBM via a Cognitive Assistant capability makes recommendations concerning what algorithms to employ based on the type of application being built. That is critical in terms of both the quality of the analytics application being built as well as the amount of time required to craft it, given the fact that advanced analytics expertise remains in short supply, says Thomas.
Given the massive amount of data an analytics application now routinely accesses, usage of public cloud resources is generally required. The challenge is figuring out a way to get as much of the underlying hybrid cloud infrastructure out of the way of the analytics as possible.