May 2012 Summaries

Cassandra is often used for storing time series data, such as application/server metrics for monitoring and capacity planning purposes. The basic data model for metric storage in Cassandra consists of a metric ID (the row key) and a collection of timestamp/value pairs (columns in a row). OpsCenter, a tool for managing and monitoring Cassandra clusters, tracks metrics about nodes and stores this data inside Cassandra before presenting it as user-friendly graphs. Metric tracking/collection is a common use case in Cassandra, with specific practices including using timestamps as column names to automatically sort data points by time, aggregating raw collected data points into more manageable sets, and preventing rows from becoming too large through expiring older metric data or sharding metric rows.

CQL3, the latest version of the Cassandra Query Language, introduces a range of enhancements beyond its primary changes of supporting composites and wide rows. These include cleaning up inconsistencies like case sensitivity in identifiers, now defaulting to case insensitive unless specified with double quotes, and improving query capabilities with strict bounds, unlike its predecessor CQL2. CQL3 also introduces the `timeuuid` type for time series data, enhancing the use of Type 1 UUIDs, and allows direct querying of system tables, improving schema management. The language forbids unintuitive queries with random partitioners, providing the `token` function for more accurate querying, and addresses the issue of range ghosts by ensuring rows returned have at least one non-primary key column set. It optimizes query performance through the CLUSTERING ORDER option, allowing efficient sorting in descending order, which is beneficial for time-series data. These changes, alongside ongoing developments, aim to make CQL3 more user-friendly and powerful, with documentation updates expected soon.

The text discusses the use of R in data analysis and its integration with other tools such as Hadoop and Cassandra. It mentions that while some data scientists prefer to integrate Hadoop with R, others want ways to get data into R and use data sampling techniques. The author demonstrates how to interact with Cassandra from R using the RJDBC module and provides an example code for accessing Cassandra data from R. Additionally, it mentions a new RCassandra package and DataStax Enterprise for accessing Hive and Cassandra from R. Overall, the text highlights the power of combining R's statistical methods with other data analysis tools.

A graph database is a software system that stores data as interconnected vertices and edges. These databases are optimized for executing graph traversal processes. The structure of graphs shares similarities with neural systems like the human brain, which can be described as networks of neurons. As graph systems scale to include more diverse data, multi-level structural understanding becomes crucial for studying graphs and designing graph systems. Neuroscience may foster an appreciation of various structural abstractions within graphs. In both cognitive neuroscience and network science, it is common to abstract away low-level connectivity patterns to identify larger functional structures. Functional motifs can be identified in real-world graphs, similar to the brain's functional areas. Graph databases are developing infrastructure capable of representing and processing complex information landscapes within a unified structure, emphasizing the importance of structural abstractions for better reasoning about graphs and designing algorithms for collective problem-solving.