Mongoose is a storage performance testing tool.
It is designed to be used for:
Mongoose is able to sustain millions of concurrent connections and
millions of operations per second.
Basically, Mongoose may be started very simply:
java -jar mongoose.jar
Mongoose has rich configuration subsystem supporting the
Item is the unit of the load progress and metrics accounting. An item may be a file,
cloud storage data object, some kind of token or something else. The
data items describe the data to use to perform a load
operation (some fixed or random size, content source, etc).
Different data may be used to perform a load on a storage. By default Mongoose uses random,
uncompressible data. An user may supply
custom data input file instead.
In contradiction to all other comparable performance testing software, Mongoose doesn't use the
threads to set up a specified concurrency level. This allows to use
unlimited concurrency levels easily.
The so called recycling feature allows to use the limited amount of items for an unlimited
test run/load job by reusing these items again and again (circularly).
A test step is an unit of a test run with its own configuration and metrics reporting.
The metrics reported by Mongoose are designed to be most useful for performance analysis.
The following metrics are accounted:
- Counts: items, bytes, elapsed and effective times.
- Rates: items per second and bytes per second.
Timing distributions for operation durations and network latencies.
There are the metrics outputs performed periodically while a load job is running and
total ones when a load job is finished.
Also, it's possible to
obtain the highest-precision metrics (for each operation).
The CRUD notation is used to describe the load operations. However, there are some extensions:
- Create may act as copying the items from the source to the destination
- Read may validate the data on the fly
- Read may be partial
- Update may act as append
Noop operation is also available.
One of the most interesting things is a data reentrancy. This allows to validate the data
read back from the storage successfully even after the data items have been randomly updated
multiple times before.
Mongoose is able to run the tests described with scenario files in the JSON format. Scenario
syntax allow to:
Mongoose is able to work in the distributed mode what allows to scale out the load
performed on a storage. In the distributed mode there's a controller which aggregates the
results from the several remote storage drivers. Storage drivers perform the actual
load on the storage.
Mongoose supports different storage types: a
filesystem or a cloud storage.
Currently, the following cloud storage APIs are supported:
It's possible to implement a custom storage driver to extend the storages support.