We were reviewing DynamoDB. We also reviewed Cosmos DB. Azure Cosmos DB was written to be the industry's first globally-distributed, multi-model fully managed database services. The latter is built to collect data from a plethora of IoT devices as well as to be handle the processing of machine learning. Today we compare the Dynamo DB and Cosmos DB.
First, Cosmos DB shows global distribution. As more and more regions are added, the same entity can be spread out over any number of regions. Dynamo DB does not offer region by region addition for the same entity. Each entity is local to a region.
Second, The criteria above also implies that storage and throughput can now be determined by more than one regions where as Dynamo DB cannot make that claim. Although both first and second are not the functional database features an application might use, it improves the availability and performance of the database to the application.
Third, spanning of multiple regions also provide multi-homing capabilities that are otherwise missed from dynamo db database.
Fourth, there is no compromise on read/write latency which is guaranteed to be lowest in Cosmos and at the 99th percentile. The same kind of metric is atleast not mentioned for Dynamo DB.
Fifth, Cosmos offers five consistency models where Dynamo offers only two of those namely strong and eventual. The availability of more consistency models allows an application to pick the behaviour of the database that best suits its persistence requirements.
Sixth, Cosmos offers support for multiple models and API. Dynamo does too. This is the flexibility that the applications want especially since they are fast becoming standards.
Seventh, Cosmos does not require maintenance. It does not maintain any schemas. Dynamo however requires indexes to be managed. This also leads to the feature of Cosmos where indexes are automatically maintained for all data.
Eighth, the service level agreements for the Cosmos are far more comprehensive than Dynamo. Although, the SLAs for NoSQL databases do not seem to have the same rigor as transactional systems, the performance and availability of the database make it more appealing.
#codingexercise
Convert multiset to distinct combinations where order does not matter:
Solution: http://ideone.com/ARkeZe
Some more http://ideone.com/lKxg5s
Also, it is possible to use a tree structure for the solution involving nested dictionaries shared yesterday.
First, Cosmos DB shows global distribution. As more and more regions are added, the same entity can be spread out over any number of regions. Dynamo DB does not offer region by region addition for the same entity. Each entity is local to a region.
Second, The criteria above also implies that storage and throughput can now be determined by more than one regions where as Dynamo DB cannot make that claim. Although both first and second are not the functional database features an application might use, it improves the availability and performance of the database to the application.
Third, spanning of multiple regions also provide multi-homing capabilities that are otherwise missed from dynamo db database.
Fourth, there is no compromise on read/write latency which is guaranteed to be lowest in Cosmos and at the 99th percentile. The same kind of metric is atleast not mentioned for Dynamo DB.
Fifth, Cosmos offers five consistency models where Dynamo offers only two of those namely strong and eventual. The availability of more consistency models allows an application to pick the behaviour of the database that best suits its persistence requirements.
Sixth, Cosmos offers support for multiple models and API. Dynamo does too. This is the flexibility that the applications want especially since they are fast becoming standards.
Seventh, Cosmos does not require maintenance. It does not maintain any schemas. Dynamo however requires indexes to be managed. This also leads to the feature of Cosmos where indexes are automatically maintained for all data.
Eighth, the service level agreements for the Cosmos are far more comprehensive than Dynamo. Although, the SLAs for NoSQL databases do not seem to have the same rigor as transactional systems, the performance and availability of the database make it more appealing.
#codingexercise
Convert multiset to distinct combinations where order does not matter:
Solution: http://ideone.com/ARkeZe
Some more http://ideone.com/lKxg5s
Also, it is possible to use a tree structure for the solution involving nested dictionaries shared yesterday.
No comments:
Post a Comment