Lindorm for Cassandra: A better foundation for cloud Cassandra applications-ApsaraDB for Cassandra

This topic describes how Alibaba Lindorm enhances the performance of Cassandra for cloud applications.

Introduction to Alibaba Lindorm

Alibaba Lindorm is a cloud-native, multi-model database service that operates at any scale. It supports low-cost storage and processing for massive amounts of data with elastic, pay-as-you-go billing. Lindorm provides several data models, including wide table, time series, search, and file. It is compatible with open standard interfaces such as HBase, Cassandra, Phoenix, OpenTSDB, Solr, and SQL. Lindorm is a preferred database for scenarios such as the Internet, IoT, Internet of Vehicles, advertising, social media, monitoring, gaming, and risk control. It is also a key database supporting Alibaba's core business.

Alibaba Lindorm is built on a cloud-native architecture that uses storage-compute disaggregation and integrated multi-model data sharing. It offers benefits such as elastic scaling, low cost, ease of use, openness, and stability. Lindorm is ideal for storing and analyzing data such as metadata, logs, bills, tags, messages, reports, dimension tables, sink tables, feed streams, user personas, device data, monitoring data, sensor data, small files, and small images. Its core capabilities include the following:

Integrated multi-model: Lindorm supports four models: wide table, time series, search, and file. It provides both unified federated queries and independent open source interfaces. Data is interoperable between models, which makes application development more agile, flexible, and efficient.
Excellent cost-effectiveness: Lindorm supports high-concurrency throughput in the tens of millions and provides millisecond-level access latency. It significantly reduces storage costs using high-density, low-cost storage media, smart hot and cold data separation, and adaptive compression.
Cloud-native elasticity: Lindorm supports independent elastic scaling for compute and storage resources. It also provides a serverless service that offers on-demand elasticity and pay-per-use billing.
Open data ecosystem: Lindorm provides easy-to-use capabilities for data exchange, processing, and subscription. It is highly compatible with systems such as MySQL, Spark, Flink, and Kafka.

Challenges of open source Cassandra

Open source Cassandra is a distributed NoSQL database based on Amazon DynamoDB and Google Bigtable. It offers advantages such as a masterless architecture, tunable consistency, and a SQL-like query language (CQL). However, Cassandra presents several challenges in practice. For example, Cassandra does not separate storage and compute. When scaling out, data must be migrated. This process is time-consuming and cannot respond quickly to sudden traffic bursts. In contrast, Lindorm uses a storage-compute disaggregation architecture, which allows for rapid elastic scaling.

Additionally, Cassandra requires periodic full data repairs across all replicas. Otherwise, issues such as "ghost keys" can occur. This repair process consumes a large amount of system resources and can affect service stability. In Lindorm, the underlying storage component handles consistency. When data is written, it is written to the specified number of replicas. This prevents data inconsistency and eliminates the need for periodic repairs.

Lindorm for Cassandra features

Higher performance: Compared to open source Cassandra, Lindorm for Cassandra delivers higher throughput and lower latency for large-scale data.
Transparent hot and cold data separation: Lindorm for Cassandra lets you configure storage media and compression algorithms for hot and cold data. This capability reduces storage costs for cold data and improves access performance for hot data by providing integrated separation with automatic data tiering.
On-demand billing: Lindorm for Cassandra is available in cluster and serverless editions. The serverless edition is fully managed and scales elastically based on business demand. Billing is based on the number of requests and the amount of data stored, which provides a simple, cost-effective, and flexible billing model.
Integrated search and wide table: Lindorm for Cassandra uses full-text indexes to accelerate queries on the wide table engine. It provides a unified access method through the CQL language.
Rich data tunnels: Lindorm for Cassandra provides a variety of data tunnels to other systems, such as search engines, online transaction databases, data warehouses, and log queues.
Additional enterprise features: Lindorm for Cassandra offers various other enterprise-grade features, as shown in the following figure. For more information about these features, see the Lindorm product homepage.