Study Notes – Databricks Cram Sheet

What’s the difference between databricks and spark?

  • Databricks is PaaS platform built on spark that offers all the additional features required to easily productionise spark into an enterprise grade integrated platform with 10-40x performance gains. Comparison is here

Is Databricks database software?

  • No – It’s a distributed calculation engine that provides an analytics, streaming, data lake and data warehouse platform across distributed nosql storage

What distributed storage can it run on?

  • AWS S3
  • Azure Data Lake Storage I think possibly even blob not sure yet
  • Hadoop

What cluster managers does it support for distributing the calculation engine?

  • YARN
  • Mesos
  • Spark – built in standalone for dev & learning

What is it implemented in?

  • Scala

What programming languages does it support?

  • Python
  • Java
  • R
  • Scala
  • SQL

What class of use could I use it for?

  • Streaming
  • SQL Analytics
  • Data Transformation (Batch or Realtime)
  • Data Provisioning into Data Warehouse or Data Lake solution
  • Deep Learning
  • Machine Learning (Batch or Realtime)
  • Graph Analysis

What core API’s does it have?

  • MLib – machine learning
  • Streaming
  • SQL
  • GraphX

Can I use 3rd party non-core API’s?

  • Yes

It’s api’s are unified but what does that mean?

  • It means code can be ported from streaming to batch with little modification; lots of work has been put in to minimise time to production, ease of development and migrate solution from a streaming to batch analytics solution for example with ease

Is it free?

  • Spark is Free Databricks is not

How can I use it?

  • Databricks has a cloud portal – there is a free trial
  • Databricks can be provisioned on AWS
  • We’ll soon be able to provision databricks in Azure – it’s on preview

What features differentiates it as a leading data platform?

  • Unified coding model gives shorter dev cycles and time to production
  • It’s PaaS – no hardware cluster to manage, create or look after and I can easily scale it
  • Has a rich collaborative development experience allowing data engineers and data scientists to work together
  • I can run data processing and querying over S3, Azure Data Lake Storage and Hadoop HDFS with:
      • Much greater performance than other distributed storage query engines
      • Automatic Index Creation
      • Automatic Caching
      • Automatic Data Compacting
      • Transactional Support


  • There is no buy into a proprietary storage format – i.e. it just sits S3 for example and I can access and manage it with other processes and tools
  • Delta (2018) transactionally incorporates new batch and/or streaming data immediately for queries – no other data platform has this