Cloud Bigtable Architecture Google Professional Data Engineer GCP

• Data is stored as a sorted map structure.
• Data is indexed at the column level as a group of
  • row key
  • a column key
  • and a timestamp
• The index is the key mapping to the column’s actual value.
• Timestamp is for versioning.
• Data operations are conducted on a per-row basis so, all columns of a row are updated simultaneously.
• Automatic data compression is also done as applicable.
• Bigtable cluster may only operate within a single zone.
• For higher availability, configure with two separate clusters, each in a different zone of the same region.

Cloud Bigtable Architecture:

• Client requests go through a front-end server
• Nodes are organized into a Cloud Bigtable cluster of a Cloud Bigtable instance
• Each node in the cluster handles a subset of the requests to the cluster.
• Add nodes to increase the number of simultaneous requests to handle and maximum throughput
• Table is sharded into blocks of contiguous rows, called tablets similar to HBase regions. Tablets are stored on Colossus, Google’s file system, in SSTable format.
• An SSTable is a ordered immutable map from keys to values, and both are byte strings.
• Tablet is associated with a specific node.
• Writes are stored in Colossus’s shared log as acknowledged
• Data is never stored in nodes themselves;
• Node have pointers to a set of tablets stored on Colossus.
• Rebalancing tablets from one node to another is very fast
• Recovery from the failure of a node is very fast
• When a Cloud Bigtable node fails, no data is lost.