Accumulo

Chapter 8. Table Design

Accumulo provides application developers with a high degree of control over data layout, which has a large effect on the performance of various access patterns. Here we discuss some table designs for various purposes and address particular issues in designing keys, values, and authorizations.

Single-Table Designs

Some applications require looking up values based on a few specific pieces of data most of the time. In these cases it is convenient to identify any hierarchies that may exist in the data and to build a single table that orders the data according to the hierarchy.

For example, if we are writing an application to store messages, such as email, we might have a hierarchy that consists of user accounts identified by unique email addresses. Within a user account we have folders, and each folder contains zero or more email messages (Figure 8-1).

In addition to natural hierarchies in the data, we also need to consider access patterns. A common query will be to access a list of messages to or from a user within a particular folder, preferably in time order from most recent to oldest.

Figure 8-1. Hierarchy in data

An example application method for fetching this data can look like the following:

listMessages(emailAddress, folder, offset, num)

where emailAddress is the user’s email address, folder indicates which mail folder to access, and offset and ...