Case Study

Data Discovery for Data Science

<< Back to Customers


Once an organization has established a process for ingesting data into a data lake, there must be an easy way discover, inspect and track datasets within the data lake. One organization’s data scientists and business analysts found it difficult to locate the datasets in their data lake and discover them based on business or technical metadata. This led to a lengthy process of having to regenerate datasets and redo the same work over and over, thereby wasting time, which required additional costly cluster capacity as well as staff resources. Some of the additional issues were:

• The time it took to discover datasets needed for ideation was days or even weeks
• Due to the difficulty of locating datasets, cluster resources were over-utilized to re-create datasets already present on the cluster
• Multiple instances of the same datasets created confusion about the most recent and authoritative copy
• IT spending was unpredictable as more data scientists and analysts were added to the team

With the Cask Data Application Platform (CDAP), all datasets generated were tracked and indexed by both technical and business metadata, allowing data scientists and data analysts to discover all datasets and inspect tags, schema and properties. They were then able to use the automatic lineage tracking capabilities to determine the source of the dataset and understand the transformations applied. Operational metadata helped the organization identify the type and frequency of processing applied on the datasets, in addition to answering the question who performed it; also, the audit data captured helped determine the freshness and activity-level of the datasets.

Benefits of Cask Solution

Rapid Time to Value
    • Time to discovering datasets on a data lake was reduced from days or weeks to minutes or hours.
    • Seamless integration between CDAP took the company from the data discovery phase to ideation and pipeline creation in minutes, which previously required hours and sometimes days.
    Lower Cost of Ownership
          • Having an easier method to discover datasets in the data lake lowered the utilization of cluster resources in terms of compute and storage.
          • Ultimately IT spending became much more predictable.
    Business Empowered with Self-Service Analytics
        • Lineage and audit capabilities allowed users to obtain authoritative answers to source, transformation and freshness of data, increasing transparency and their trust in the quality and nature of datasets.
        • Collaboration between data engineers, scientists and analysts improved.
    Download Case Study