The Object File System (OFS) integrates object storage into VantageCloud Lake to provide some of the attributes of the Teradata File System. The Object File System is a private and dedicated space which stores data and objects such as user tables managed by the Analytics Database.
- The Object File System provides high durability and availability, so fallback is not needed. This not only helps in reducing space usage by half, but also reduces the maintenance cost of DML processing.
- Data can be accessed from the primary cluster or a compute cluster.
- Storage is not statically associated with an AMP. Therefore, as AMPs are added to the system, the rows need not be copied to a new map.
- Data on an Object File System is encrypted.
Using the Object File System for storage provides higher availability because all compute clusters have access to the Object File System. If one compute cluster goes down, other compute clusters (or a new compute cluster spun up for this purpose) can still access the data on an Object File System.
Tables in the Object File System allow one writer to INSERT/UPDATE/ DELETE rows into/from the table without blocking one or more readers.
Older versions of a table are retained for a period of time, which allows readers to see the state of a table at some previous point in time.
The Parquet format is used to store and process column-partitioned tables efficiently in the Object File System.
- Object storage stores data in objects. There is a high latency and a very wide distribution of latency when accessing objects in object storage. This implies more end user involvement in planning the loading of data into tables to determine the best storage option for different workloads.
- Tables in an Object File System do not have a primary index. Therefore, these tables are not optimal for single-row access and several types of join plans cannot be used.