2026_COMP_MTech_MTCST 204(B): Big Data Analysis

Download the ppt: https://www.dropbox.com/scl/fi/sqtdndeiw297gbe17ohtr/07_Cassandra.pptx?rlkey=ngf7igwqqajk5l2eokf1o9486&dl=0

• What is Gossip Protocal?

Gossip protocol is a process of computer peer-to-peer communication that allows nodes in a distributed cluster to talk to one another.
Key functions and characteristics of the gossip protocol include:

• Sharing Metadata: Nodes use this protocol to share and discover information about the metadata of the network and the cluster.

• Determining Data Responsibility: Through gossip, nodes communicate about which specific node is responsible for which token ranges (data partitions).

• Request Coordination: When a client request hits any node in the cluster (acting as a coordinator), that node uses the information gained through the gossip protocol to identify the correct node responsible for that data and forwards the request accordingly.

• Masterless Architecture Support: Because Cassandra has a masterless architecture where every node is equal, the gossip protocol is essential for ensuring all nodes stay updated on the status and location of their peers without a central authority

---------------------------------------------------------------------------------------------------------------------
What is Query-first modeling?
Query-first modeling is a database design methodology used in Apache Cassandra where tables are designed specifically to satisfy the exact queries an application will perform, rather than modeling based on entities or normalization as is common in SQL.
According to the sources, this approach involves the following key principles:

• Focus on Fetching over Storage: Instead of thinking about how to store data efficiently to save space, developers must think about how they will fetch the data and what the specific use cases are. This ensures that when a query is fired, it hits a node that can return the data instantly without complex processing.

• Denormalization: Unlike relational databases (SQL) that use normalization to avoid repeating data, query-first modeling favors denormalization. This means data is purposefully repeated across different tables so that all information needed for a specific query is available in one place, eliminating the need for time-consuming JOIN operations.

• Optimization for Speed: The primary goal of this model is to retrieve data in a fraction of a second. In a world of "Big Data," joining multiple huge tables is considered a "weakest link" because it is too slow for real-time transactional workloads.

• Primary Key and Partition Design: In this model, the primary key (specifically the partition key) is chosen based on how the application filters data. For example, if a website frequently searches for users based on their location, the table would be designed with "location" as the partition key to ensure all related data is stored together on the same node for fast retrieval.

• Clustering Keys for Sorting: After the partition key is determined, clustering keys are used to define how the data is sorted within that partition, further optimizing the specific query's output.