Data lake is never used as a store (persistence mechanism) for transactional systems. However, many business services exposed in the Data Lake can be used by OLTP systems to cater many use case requirements. Business services typically consume multiple data services to provide a business capability, while data services operate at data level, ensuring that the data is exposed in the its most natural form from the data platform, without any influence of business logic or business processing. Business logic and business processing should happen at Business services level so that we can achieve a loosely coupled services ecosystem while keeping data services at its purest forms.

One of the examples that we could think of around Business Service is as follows:

Consider you have an OLTP application used for selling a product. You have already pumped good amount of data in your Data Lake and have analytic logic build in for finding recommended products for a customer. This product recommendation could be exposed as a business endpoint (REST over HTTP) and could be sued by your OLTP application to show product recommendation when customer is in your website. This recommendation analysis in Data Lake can make use of really old customer behaviour data (this data being old would have already gone away from production datastore, stored in our Lambda Batch Layer) and new data (present transaction data, in our Lambda Speed Layer, flown in real-time from the OLTP application).