Beyond the Limitations of Cloud Platforms

Distributed Transactions and Scaling

Distributed two-phase locking easily blocks when one node is unavailable.
Distributed two-phase locking protocol is a source of significant communication overhead.
Deadlocks in homogenous distributed environment are more frequent and considerably more difficult to detect.
With distributed transactions, the releasing of locks is subject to the two-phase rule for the scope of the entire global transaction.
Consistency of distributed transactions is dependent on strict serialization with distributed two-phase locking, which excludes the scaling as viable option.

The elastic scaling of Cloud computing platforms:

Cannot scale distributed relational DB transactions.
Do not support stateful sessions or sessions' affinity impairs the elastic load balancing.
Cannot create new instances in responsive manner. It takes minutes to create a new instance and seconds to revive a stopped one.

The elastic scaling of Transactum Engine:

Does not use distributed two-phase locking.
Uses invariants to make operations on distributed data follow certain access patterns.
Supports stateful sessions. The elasticity of load balancing and elasticity of scaling are not affected by sessions' affinity.
Creates a new instance in less than 100ms and revives a recycled instance in less than 10ms.