Budget constraints and cost are the top reasons why many companies are shying away from deploying big data, according to a study performed by Deloitte. It can be hard to justify investing in new IT infrastructure to process large amounts of data, especially if the business does not yet havean immediate business case.

Processing big data workloads is different than processing typical enterprise application workloads. Big data workloads are processed in parallel, instead of sequentially. IT typically prioritizes business critical workloads and schedules lower priority jobs in batches at night or when there is excess capacity. With big data analytics, many use cases must be run in real-time for live analysis and reaction. This forces IT to change data center policies and learn new tools to create, manage and monitor these new workloads.

Companies need to choose the right type of infrastructure to run their applications and data. Procuring hardware takes time. Going into the cloud may be great for a proof of concept, but it carries the risk of platform lock-in, comes with security concerns and incurs tremendous cost at scale. Companies must also decide which Hadoop distribution to select, with Cloudera, Hortonworks, MAPR, and Pivotal all offering competing (and incompatible) architectures. There are many decisions that, once made, make it difficult for a company to pivot later so many companies just delay having the big data conversation.