November 2012 Summaries

The text discusses three methods for moving data between DataStax Enterprise (DSE) and Cassandra, and other sources. Firstly, the COPY command in Cassandra 1.1 and higher allows loading flat file data into Cassandra or writing data out to OS files. Secondly, Sqoop is supported by DSE 2.0 and higher for transferring data between RDBMS and Hadoop, including moving data directly into Cassandra. Thirdly, ETL solutions like Jaspersoft, Pentaho, and Talend can be used for more sophisticated data movement situations, offering transformation routines and additional features.

The first version of Cassandra Query Language (CQL) was introduced in Cassandra 0.8 as an alternative to Thrift API. CQL has been using Thrift as a network transport due to its convenience and speed, but it is not tied to Thrift for the transport. In Cassandra 1.2, a new binary protocol will be introduced that is specifically tailored for CQL3. This protocol offers features such as asynchronicity and notifications. To use this new protocol, users need to have a version of Cassandra 1.2, activate the binary protocol server in cassandra.yaml, and use a client driver that supports the new protocol. DataStax has open-sourced a Java Driver for this purpose.

Cassandra automatically saves trace sessions to the system_traces keyspace for reference, which expires after 24 hours. The system_traces keyspace contains two tables: sessions and events. Cqlsh's rendering of a traced request omits request parameters and event thread, but these can be useful when probabilistic tracing is enabled. Activity is a simple text field that may change in future Cassandra releases. Nodetool settraceprobability allows coordinators to trace a proportion of all requests it handles, which can help identify intermittent query slowness. However, tracing requests requires significant resources and should be used judiciously, starting with a small fraction like 0.001 and increasing only if necessary.

Cassandra, a distributed database system, has moved beyond its original design and now supports more complex read operations. To help users understand what's happening under the hood, request tracing has been added to version 1.2. Tracing can be used to diagnose performance problems and optimize data models. For example, appropriate indexes are needed for efficient queries, and using Cassandra as a durable queue may require rethinking the data model due to its log-structured storage engine. Cassandra 1.2 Beta 2 is available for testing, with the final release expected before the end of the year.

DataStax Enterprise leverages Cassandra's innate datacenter concepts, allowing multiple workloads to run across multiple datacenters without ETL processes or manual operations. This enables a natural distribution of data from real-time datacenters to near real-time analytics and search datacenters. Implementing datacenters as divisions between varying workloads allows for efficient handling of writes and updates, with the option to use different datacenters as live backups that can quickly be used as fallback clusters. Additionally, data can be housed in different datacenters based on user location to provide more responsive exchange while maintaining low operational costs by keeping all datacenters within the same cluster.