Lambda architectures consists of an arrangement of data and processing technologies intended to serve both regular batch processing activities and near real-time streaming activities, in a way that allows results from both to be combined into a single view. The following descriptions and diagram highlight the conceptual components that make up the three layers to the architecture:

• Batch: This layer keeps the master dataset and regular batch processing of the data to create views to support analytics, dashboards, and reporting linked to the Serving layer. The master dataset is immutable, meaning the data does not change once incorporated, and it is appended with new data as time goes on. A long-running Hadoop cluster would be an example.
• Speed: This layer is focused on recent data only (the streaming data). It handles requests that have low-latency requirements in terms of recency of data. Algorithms need to be fast and incremental to operate in the speed layer. Amazon Kinesis would be an example technology that operates in the speed layer.
• Serving: This layer indexes batch views so they can be served up with low latency. This includes batch views from both the Batch and Streaming layers.