HomeAncelus TopicsAncelus Scalability

Ancelus Scalability

The Ancelus database delivers unprecedented performance, constant at any size to 32 exabytes.

Smaller Dataset

Ancelus drives improved performance by eliminating most of the duplicated data in structured data storage.  This reduces the dataset size by 85 to 98%.

Massively Parallel Access, Locking

The parallel read/write access is enhanced by the column level locking. This means that parallel writes to different columns in the same table do not produce a collision.

When collisions occur, a special implementation of spinlocks reduces the lock/unlock cycle to 20 nanosecond.  In high traffic implementations there is a facility for dynamic priority inversion to avoid the deadlock bugaboo. 

But the real secret of Ancelus collision management is simply not to do it.  If a R/W process completes in less than 100 nanoseconds, the chance of a collision that requires use of our high speed locking utility shrinks to very near zero. 

Massively Parallel Hardware

When deployed in multi-server environments Ancelus continues the quest for speed.  Queries are organized in the database so that the extraction only delivers the result. This reduces the demand for backplane interconnect bandwidth.  The client receives only the interesting data. The heavy lifting is already done in Ancelus database.  Since query paths are stored in the data dictionary, the logical work is done before the extract begins.

Because there is no table level or database level locking in Ancelus, the database is inherently multi-threaded and can run multiple concurrent operations on parallel cores, CPUs and servers.

The foundation of this exceptional performance is the Ancelus algorithim.  It only takes 4 cpu operations to resolve the physical address even when the physical location is on a different server.


Go to top View Our Stats