---

Characteristics of NOSQL systems

---

• Characteristics of NOSQL systems
  • Emphasis in NOSQL systems:

    High performance (due to a large user population)              Scalability Availablity

  ---

  ---

  ---

  ---

• Scalability
  • Scalability:

    Scalability = the ability of a system to handle a growing amount of work

    ---

    ---

  • 2 ways to achieve scalability:

    Vertical scalability: Handle more load by using faster processors         Horizontal scalability: Handle more load by using more processors

    ---

    ---

    ---

  • NOSQL systems are always horizontally scalable:

    Processing/storage capacity in NOSQL systems are increased by adding more processing/storage nodes (I.e., it does not rely on using faster hardware (= vertical scalability))

    ---

    ---

  • The horizontal scalability of the NOSQL system is employed while the NOSQL system is operational !!!

    You do not need to shut down the NOSQL system to add more nodes !!!

  ---

  ---

  ---

  ---

• Availability

  • Availability:

    Availability = the likelihood that the system is operational Many applications that use NOSQL system must operate continuously

    ---

    ---

  • Techniques to achieve high availability:

    Replication !!!         Replication will eliminate any single point failures          (At a high cost)

    ---

    ---

  • Replication techniques (for data storage):

    Master-slave replication: One of the copies is the master copy All write operations must apply to the master copy The slave copies will eventually receive the updates Read operations usually can access any data copy for performance Master-master replication: Allows read/write to any replica (copy) The write operation includes a time stamp A reconciliation algorithm propagates the latest update to all replicas eventually

  ---

  ---

  ---

  ---

• High performance
  • Performance:

    The performance of a system is often measured by: the response time of an action/request

    ---

    ---

  • Methods to achieve higher performance:

    Lower the processing load per node How: Distribute the data/requests over different nodes        Increase the level of parallellism How: Relaxing the data consistency constraint !!! (This will reduce the number of synchronization bottle necks and more processes can run simultaneously !!!)

  ---

  ---

  ---

  ---

• Technique to reduce the processing load per node
  • Sharding (= horizontal partitioning)

    ---

    ---

  • How the data is stored in sharding:

    Data is partitioned horizontall into "shards" Each "shard is stored at a different node Graphically:

    ---

    ---

    ---

  • Load distribution using sharding:

    Processing is distributed over multiple nodes: Without sharding, a single node must process the entire data object: With sharding, multiple node can share the processing load:

  ---

  ---

  ---

  ---

• Sharding combined with data replication
  • Availability and higher performance can be achieve by combining sharding and data replication:

  ---

  ---

  ---

  ---

• Commonly used sharding techniques
  • Commonly used sharding techniques:

    Hashing Range partitioning

    ---

    ---

    ---

  • Hashing-based sharding:

    Used in key-value NOSQL systems Data is sharded using the key:

  • Note:

    This approach is similar to the tuple storage in parallel database processing:

    ---

    ---

    ---

    ---

  • Range-based sharding:

    Data is sharded based on pre-defined ranges of key values:

    Note:

    Range-based sharding can handle range-based queries !!!

  ---

  ---

  ---

  ---

• Increasing parallelism in NOSQL systems
  • NOSQL system adopt relaxed data consistency constraints:

    NOSQL systems does not enforce strict serializability as the consistency condition

    ---

    ---

  • "Eventual" consistency:

    NOSQL uses a less stringent condition called: "eventual" consistency" An update is applied to one copy without locking the other copies: So: the data can be temporally inconsistent (see above figure) !!!!