January 2012 Summaries

The text discusses the comparison of operating systems, specifically Linux, Windows, and Mac, in terms of their performance with databases such as Oracle, SQL Server, MySQL, and PostgreSQL. It mentions a case where a Windows configuration outperformed an HPUX system for a major trading system at a large company. The author also talks about the new bundled installer for Apache Cassandra on Windows, which includes utilities like CQL interface and DataStax OpsCenter community edition. The MSI installer is designed primarily for single Windows installs on workstations and laptops but can be tweaked to create multi-node setups. Currently, the package supports Windows 7 and Windows 2008 Server, both 32 and 64-bit, for development work only, with production support coming soon. The author encourages users to download and provide feedback on the new installer.

In response to user requests, DataStax has released OpsCenter on Mac. Users can now download and run the software on their Apple devices. The installation process for both Cassandra and OpsCenter is detailed in a new article. Currently, support for OpsCenter on Mac (both Community and Enterprise versions) is limited to development purposes only. Feedback from users is welcome as DataStax continues to improve its offerings.

The text provides a step-by-step guide to installing Apache Cassandra and DataStax OpsCenter on Mac OS X. It explains how to download and set up Cassandra, its utilities, and use DataStax OpsCenter, which is a browser-based visual management and monitoring tool for Cassandra. The text also covers the installation of DataStax Community Edition from Planet Cassandra, using the curl command on Mac to directly download files, creating keyspaces and column families in Cassandra, and querying data using CQL (Cassandra Query Language). It concludes with instructions for installing DataStax OpsCenter on a Mac.

Cassandra 1.0 introduced two compaction strategies: SizeTieredCompactionStrategy and LeveledCompactionStrategy. Choosing between them can be challenging due to the dependence on application behavior and requirements. Testing compaction strategies is difficult as it requires re-compacting all data, which takes a long time and only allows static analysis. Cassandra 1.1 introduced write survey mode, allowing live traffic sampling to a test node without significantly impacting the existing cluster. This feature can be exploited for comparing compaction strategies by following specific steps. The ability to change compaction strategy via JMX ensures it's only changed on the test node. In the future, testing compression settings using this method is also planned.

This article provides a step-by-step guide to setting up and monitoring a multi-node Apache Cassandra cluster on Linux using Ubuntu 11.04 machines. The process involves confirming prerequisites, downloading the DataStax Community Edition bundle, generating tokens for each machine in the cluster, configuring startup parameters, starting nodes, installing DataStax OpsCenter, and monitoring the cluster. The guide also mentions that the same process can be applied to other Linux platforms such as Red Hat, CentOS, and Debian.

Amazon has introduced DynamoDB, a hosted database for its AWS platform. The new service shares similarities with Apache Cassandra, as it also achieves high scalability using many of the same techniques. Both systems can handle millions of operations per second and have multi-datacenter availability features. However, DynamoDB lacks cross-region replication capabilities and offers a simpler data model compared to Cassandra's more powerful ColumnFamily design. Additionally, while both support Hadoop integration for analytical queries, it is unclear if DynamoDB can partition workloads like Cassandra does. Lastly, Cassandra has an advantage in real-world features such as backup and log-structured storage engine, which allows full and incremental backups without impacting performance.

Composite columns in Apache Cassandra are useful for adapting some models and providing new indexing functionality, particularly for time series data. They help in efficiently accessing data by minimizing disk seeks and reducing overhead associated with large numbers of skinny rows. Composite comparators can be thought of as a comparator composed of several other types of comparators. Two forms of composite comparators are CompositeType and DynamicCompositeType, with the former being discussed in this text. The use of composite columns is demonstrated through an example involving timezone data for major cities in the United States. CompositeDataLoader can be used as a model for application-level parallelized bulk loading with the Hector API, while CompositeQuery class makes use of an auto-paging feature built into Hector to provide clean iteration semantics back up to the caller.

Eventual consistency is a strategy used by distributed systems to improve query and update latencies and provide stronger availability than possible otherwise. It allows users to specify parameters such as replication factor (N), number of nodes for write success (W), and number of nodes for read operation (R). By varying these parameters, one can obtain different properties of availability, consistency, reliability, and speed. The interactive demo provided by the EECS department at UC Berkeley helps users determine their system's optimal configuration based on latency distributions and other factors.

This article provides a beginner's guide to using Apache Cassandra, a highly scalable distributed database system. It covers installation, creating keyspaces and column families, inserting data, querying data, and the use of secondary indexes. The Cassandra Query Language (CQL) is introduced as a SQL-like language for managing objects in Cassandra. The article also provides resources for further learning about Cassandra architecture, CQL, and multi-node cluster setup.