Database Reference
In-Depth Information
as VMware and Oracle. To increase adoption, many of them, now have
a Community Edition that can be used without charge for evaluation
purposes. One such product is Hazelcast.
They are primarily designed to provide a distributed view of data that
behaves as if it was simply being held in the application's main memory.
This tends to be programming language specific, with Java being a favorite,
and implements distributed versions of the native collections such as
List
and
Map
interfaces for Java.
Choosing a Technology
When persistence technologies are discussed there are inevitably arguments
over which technology is best. Like arguments about which text editor or
integrated development environment is “best,” these arguments are
entertaining but ultimately meaningless. There are really only technologies
that work well in a given environment for a given application.
Developers experienced with a specific environment may be able to
successfully generalize that experience to other applications, which may
make it the best choice simply because it offers the shortest time to market
even though another technology's characteristics would have been an
objectively better match for the application. Drawn from experiences good
and bad, this section provides some rules of thumb when considering a
particular technology.
Key-Value Stores
Key-value persistence approaches, like Redis, work best when the
aggregation structure is well known in advance. They also generally require
that the queries are fairly simple because this is managed entirely by the
client side. This makes them most appropriate for applications with limited
interaction, such as dashboard applications. Building a simple dashboard
with Redis is discussed in Chapter 7.
Document Stores
By design, document stores excel when there are a large variety of metrics to
store with a natural grouping. For example, a customer might have a large
number of customizable metrics that are specific to his business needs.
