LibDAQ: The Data AcQuisition Library
Overview
LibDAQ is a pluggable abstraction layer for interacting with a data source (traditionally a network interface or network data plane). Applications using LibDAQ use the library API defined in daq.h to load, configure, and interact with pluggable DAQ modules.
DAQ Modules
Each DAQ module implements some or all parts of the DAQ module API depending on its type and capabilities. There are two main classes of DAQ modules: base modules and wrapper modules. Base modules provide a full-fledged and independently usable implementation of the DAQ module API, while wrapper modules provide a subset of the API that is applied in a decorator pattern when combined with a base module.
DAQ Instances
A DAQ instance is an instantiation of a DAQ configuration that contains exactly one base module and zero or more wrapper modules. The wrapper modules are layered on top of the base module in LIFO order (sometimes referred to as a "module stack") and the overrides present in the wrapper module API implementations will be resolved from the top down. The basic life cycle of a DAQ instance once a configuration is defined is as such: instantiation, starting, stopping, and destruction (instantiate => start => stop => destroy).
DAQ Messages
At its core, LibDAQ is about receiving and processing data. The fundamental unit used for passing data from a DAQ instance up to the application is the DAQ message. Each message is composed of three main components: a type, a header, and some data. For example, the message used to convey a packet is of the DAQ Packet Message type with an associated DAQ Packet Header and the actual packet data itself. Messages are received in vectors/batches from the DAQ instance and finalized individually. The vector of messages received is free to be heterogeneous based entirely on what the DAQ instance happens to yield.
The basic main loop for any simple LibDAQ program will look something like this:
DAQ_Msg_h msgs[16];
DAQ_RecvStatus rstat;
unsigned num_recv = daq_instance_msg_receive(instance, 16, msgs, &rstat);
<Check the receive status in rstat for error conditions and bail>
for (unsigned idx = 0; idx < num_recv; idx++)
{
DAQ_Msg_h msg = ctxt->msgs[idx];
<Perform work on the DAQ message that results in a verdict>
daq_instance_msg_finalize(instance, msg, DAQ_VERDICT_*);
}
Repeated inside of a loop until an error or otherwise terminal condition is met.
DAQ messages may be finalized in any order and even finalized many receive calls later. Once a message is finalized, the message handle should be considered invalid by the application, as well as any and all of its data.
This is a large departure from the way LibDAQ 2.x worked. Previously, packets were received via a callback from a looping acquisition function and a verdict for each packet was the required return value from the callback. Once the callback returned, any ownership or control of the packet was surrendered back to the DAQ instance. The new paradigm allows for far more control by the application, with the responsibility that comes along with that power.
DAQ IOCTLs
DAQ IOCTLs (input/output controls) represent a semi-generic method for communicating with special functionality in the DAQ modules and the data planes they are abstracting. There are a set of first-class IOCTLs defined in the API, while IOCTL command IDs between 1025 and 65535 can be used for custom IOCTLs. Each IOCTL takes a pointer to an argument and a length and beyond that is generally free-form. An example of a first-class DAQ IOCTL would be the CREATE_EXPECTED_FLOW command, which will request that the backing data source set up the expectation for a new flow based on a potentially complex set of features to identify the expected flow with.
Although the majority of them currently do, there is no requirement that IOCTL operations pertain to a particular DAQ message or flow.
For those familiar with LibDAQ 2.x, the IOCTL functionality has subsumed the previously separate Modify Flow, Query Flow, Data Plane Add Data Channel, and Get Device Index functions.
Build and Install
LibDAQ is a standard autotools project and builds and installs as such:
./configure
make
make install
If building from git, you will need to do the following to generate the configure script prior to running the steps above:
./bootstrap
This will build and install both the library and modules.
When the DAQ library is built, both static and dynamic flavors will be generated. The various DAQ modules will be built if the requisite headers and libraries are available. You can disable individual modules, etc. with options to configure. For the complete list of configure options, run:
./configure --help