WiredTiger Data Store
WiredTiger’s storage engine is optimized for high-throughput, big data applications. It can be configured for write-optimized (row-store) or read-optimized (column-store) access, as well as a hybrid of both.
In some applications, only a fraction of each row is useful, so only the columns required for each operation will be accessed. Column storage lends itself to compression because similar data is physically contiguous. Prefix compression, run-length encoding and Huffman coding can be layered over pluggable block compression to pack many records into a single I/O.
WiredTiger separates the on-disk and in-memory representation of data, leading to a simpler, more compact file format and large block I/O tailored for modern storage systems.
Lock-free data structures are used to manage data in memory, minimizing contention between threads. A simple lightweight API, customizable with pluggable extensions, separates data representation from application logic.
WiredTiger offers substantial performance gains on existing hardware, allowing further scale up before needing to scale out. It is a solid foundation for a wide variety of data-intensive applications tailored to different problem domains.