NoSQL Final Exam With 100% Correct Answers

NoSQL Final Exam With 100% Correct AnswersWhat motivated the creation of NoSQL databases? - ANSWER 1) Impedance mismatch: differences between relational models and data structures used by applications that interact with databases 2) Web services: shift to web services (using HTTP) in early 2000's, need more flexibility in data structure 3) Big data: rapid data expansion -- either scale up or scale out 4) Clustered computing: RDBMS not optimized for clusters, ACID transactions degrade performance at scale How do NoSQL data models address impedance mismatch? - ANSWER Eliminate the need for object-relational mapping, simplify database interactions, and there is less code to write / maintain Store data the way you want to retrieve it Explain horizontal vs. vertical scaling - ANSWER Vertical scaling (scaling up): bigger machines, more expensive and ultimately reaches upper limits Horizontal scaling (scaling out): add more machines to cluster, cheaper at scale and more resilient and reliable General defining characteristics of NoSQL - ANSWER 1) Do not use relational model 2) Do not use SQL as query language 3) "schemaless" Page 2 of 10 4) Mostly designed for clusters 5) Usually open-source 6) specialized for specific use cases 7) Developed in early 2000's and on What is aggregate orientation? - ANSWER Related objects are grouped together and treated as one unit We communicate with the database in terms of aggregates Clusters use aggregates as units of data distribution What are the advantages of dynamic schema? - ANSWER Different records can have different structures, not as rigid In RDBMS, we must define structure a priori and take database down to make changes State the CAP theorem - ANSWER A distributed computer system cannot simultaneously provide : (1) consistency, (2) availability, and (3) partition tolerance Define consistency - ANSWER The client perceives that a set of operations has occurred all at once All clients see the same data "Atomicity" and "Consistency" from ACID Define availability - ANSWER Every request received by a non-failing node must result in a response Page 3 of 10 Every request receives a response about whether it was successful or failed If you can talk to a node in a cluster, it can read / write data Partition tolerance - ANSWER Operations will complete, even if individual machines are unavailable System can survive communication breakages in the cluster What do NoSQL database trade off in the CAP theorem? - ANSWER They sacrifice Consistency to achieve higher availability What do relational database trade off in the CAP theorem? - ANSWER They maintain high consistency at the expense of availability How is Consistency measured? - ANSWER Many kind of consistency measurements (between users, business logic, single request etc.) Measured in degrees 1) multi-table/ multi-operation ACID guarantees 2) strict consistency for single operations 3) eventual consistency 4) weak consistency Define Strong, Eventual and Weak Consistency - ANSWER Strong: All data are consistent immediately after the transaction completes Eventual: eventually all accesses to a value will return the latest version Page 4 of 10 Weak: no guarantees that system will become consistent How is availability measured? - ANSWER Typically defined as the percentage of time that the system is available Tuning consistency - ANSWER N: replication factor - how many copies will be stored W: number to write before operation is complete R: number to read during a read operation Define sharding - ANSWER Putting different data on separate nodes to improve DB performance this can also be done in RDBMS, but ACID transactions become difficult to guarantee • some support auto-sharding Define replication - ANSWER Putting multiple copies of data on separate nodes Master-slave: master can write, everyone can read • Redis, MongoDB, Neo4j Peer-to-peer: all nodes are equal and can read / write • Cassandra Explain range-based sharding and hash-based sharding - ANSWER Range-based: define a sharding key for the data-usually numeric or alphabetic, then assign chunk of memory based on the key; do this when you have a good reason Hash-based: pass shard key through hash function, that will assign the data to chunk of data What are the costs of replicating data? - ANSWER More storage required, we probably have to relax consistency guarantees for the system to be efficient at scale Page 5 of 10 Load balancing Schema changes No 'joins' / 'group by' across shards Quorum examples - ANSWER N W 1 : more than one node must participate in a write W N / 2 : no simultaneous updates W + R N : newest version of record is part of read operation Main data structure underlying key-value stores - ANSWER Hash tables, with more complex structures built on top How is data organized in key-value database? - ANSWER Values are unstructured from the point of view of the DB Various DB's can offer external means of declaring data types Use of key-value databases - ANSWER Web-session data Shopping carts User profiles Twitter (Redis)