---

The "CAP" Theorem

---

• Desirable properties in distributed systems
  • Ideally, we want a distibuted system to have all of these properties:

    Strong Consistency Ideally: consistency defined as serializability High availability Ideally: the distributed system is always available to process user requests Tolerant to (network) partitioning: Network failure can partition a distributed system into 2 or more operational systems Nodes within the same partition can communicate with each other Nodes in different partition cannot communicate with each other If both operational halves are allowed to operate, the result may be inconsistent !!! Ideally, we want all partitions to remain operational.

  ---

  ---

  ---

  ---

• The "CAP" Theorem
  • The CAP Theorem:

    It is impossible for a distributed system to provide: Strong Consistency High Availability Tolerance to Partitioning at the same time

    ---

    ---

    ---

  • Consequence of the CAP Theorem:

    The designer of a distributed system must choose: 2 of the desirable properties that he/she want to have in his/her system....

  ---

  ---

  ---

  ---

• NOSQL and the CAP Theorem
  • NOSQL systems always dumps:

    the Strong consistency property....

    and uses the following less stringent condition:

    Evential consistency