Batch layer | redundant and distributed processing. It aims at its data being accurate. | Hadoop, Snowflake, Redshift, Synapse and Big Query |
Speed layer | procesing real time data faster than batch layer, but with less accuracy. As soon as batch layer is finsihed with its calculation, the results of the batch layer are merged into the speed layer. | Apache Storm, Spark, SQLstream, Azure Stream Analytics, Azure Cosmos DB |
Serving layer | Stores results of speed and batch layer to be used for example for ad-hoc queries. | Apache Cassandra, HBase, Hive, Impala; Elasticsearch, Azure Cosmos DB, MongoDB, VoltDB, Elephant DB, SAP HANA |