aboutsummaryrefslogtreecommitdiffstats
path: root/PyPI/MLRsearch/README.rst
diff options
context:
space:
mode:
Diffstat (limited to 'PyPI/MLRsearch/README.rst')
-rw-r--r--PyPI/MLRsearch/README.rst118
1 files changed, 114 insertions, 4 deletions
diff --git a/PyPI/MLRsearch/README.rst b/PyPI/MLRsearch/README.rst
index 8439fe71a1..92c2afe484 100644
--- a/PyPI/MLRsearch/README.rst
+++ b/PyPI/MLRsearch/README.rst
@@ -16,6 +16,11 @@ is only a symlink to the original place of tightly coupled CSIT code.
Change log
----------
+1.0.0: Logic improvements, independent selectors, exceed ratio support,
+better width rounding, conditional throughput as output.
+Implementation relies more on dataclasses, code split into smaller files.
+API changed considerably, mainly to avoid long argument lists.
+
0.4.0: Considarable logic improvements, more than two target ratios supported.
API is not backward compatible with previous versions.
@@ -28,17 +33,122 @@ API is not backward compatible with previous versions.
Usage
-----
-TODO.
+High level description
+______________________
+
+A complete application capable of testing performance using MLRsearch
+consists of three layers: Manager, Controller and Measurer.
+This library provides an implementation for the Controller only,
+including all the classes needed to define API between Controller
+and other two components.
+
+Users are supposed to implement the whole Manager layer,
+and also implement the Measurer layer.
+The Measurer instance is injected as a parameter
+when the manager calls the controller instance.
+
+The purpose of Measurer instance is to perform one trial measurement.
+Upon invocation of measure() method, the controller only specifies
+the intended duration and the intended load for the trial.
+The call is done using keyword arguments, so the signature has to be:
+
+.. code-block:: python3
+
+ def measure(self, intended_duration, intended_load):
+
+Usually, the trial measurement process also needs other values,
+collectively caller a traffic profile. User (the manager instance)
+is responsible for initiating the measurer instance accordingly.
+Also, the manager is supposed to set up SUT, traffic generator,
+and any other component that can affect the result.
+
+For specific input and output objects see the example below.
+
+Example
+_______
+
+This is a minimal example showing every configuration attribute.
+The measurer does not interact with any real SUT,
+it simulates a SUT that is able to forward exactly one million packets
+per second (unidirectional traffic only),
+not one packet more (fully deterministic).
+In these conditions, the conditional throughput for PDR
+happens to be accurate within one packet per second.
+
+This is the screen capture of interactive python interpreter
+(wrapped so long lines are readable):
+
+.. code-block:: python3
+
+ >>> import dataclasses
+ >>> from MLRsearch import (
+ ... AbstractMeasurer, Config, MeasurementResult,
+ ... MultipleLossRatioSearch, SearchGoal,
+ ... )
+ >>>
+ >>> class Hard1MppsMeasurer(AbstractMeasurer):
+ ... def measure(self, intended_duration, intended_load):
+ ... sent = int(intended_duration * intended_load)
+ ... received = min(sent, int(intended_duration * 1e6))
+ ... return MeasurementResult(
+ ... intended_duration=intended_duration,
+ ... intended_load=intended_load,
+ ... offered_count=sent,
+ ... forwarding_count=received,
+ ... )
+ ...
+ >>> def print_dot(_):
+ ... print(".", end="")
+ ...
+ >>> ndr_goal = SearchGoal(
+ ... loss_ratio=0.0,
+ ... exceed_ratio=0.005,
+ ... relative_width=0.005,
+ ... initial_trial_duration=1.0,
+ ... final_trial_duration=1.0,
+ ... duration_sum=61.0,
+ ... preceding_targets=2,
+ ... expansion_coefficient=2,
+ ... )
+ >>> pdr_goal = dataclasses.replace(ndr_goal, loss_ratio=0.005)
+ >>> config = Config(
+ ... goals=[ndr_goal, pdr_goal],
+ ... min_load=1e3,
+ ... max_load=1e9,
+ ... search_duration_max=1.0,
+ ... warmup_duration=None,
+ ... )
+ >>> controller = MultipleLossRatioSearch(config=config)
+ >>> result = controller.search(measurer=Hard1MppsMeasurer(), debug=print_dot)
+ ....................................................................................
+ ....................................................................................
+ ....................................................................................
+ ....................................................................................
+ ....................................................................................
+ ....................................................................................
+ ...>>> print(result)
+ {SearchGoal(loss_ratio=0.0, exceed_ratio=0.005, relative_width=0.005, initial_trial_
+ duration=1.0, final_trial_duration=1.0, duration_sum=61.0, preceding_targets=2, expa
+ nsion_coefficient=2): fl=997497.6029392382,s=(gl=61.0,bl=0.0,gs=0.0,bs=0.0), SearchG
+ oal(loss_ratio=0.005, exceed_ratio=0.005, relative_width=0.005, initial_trial_durati
+ on=1.0, final_trial_duration=1.0, duration_sum=61.0, preceding_targets=2, expansion_
+ coefficient=2): fl=1002508.6747611101,s=(gl=61.0,bl=0.0,gs=0.0,bs=0.0)}
+ >>> print(f"NDR conditional throughput: {float(result[ndr_goal].conditional_throughp
+ ut)}")
+ NDR conditional throughput: 997497.6029392382
+ >>> print(f"PDR conditional throughput: {float(result[pdr_goal].conditional_throughp
+ ut)}")
+ PDR conditional throughput: 1000000.6730730429
Operation logic
---------------
-The latest published `IETF draft`_ describes logic of version 0.3,
-version 0.4 logic will be descibed in next draft version.
+The currently published `IETF draft`_ describes the logic of version 0.4,
+the logic of version 1.0 will be descibed better in the next draft version (-05).
.. _CSIT: https://wiki.fd.io/view/CSIT
.. _fd.io: https://fd.io/
.. _LFN: https://www.linuxfoundation.org/projects/networking/
.. _PyPI: https://pypi.org/project/MLRsearch/
.. _directory: https://gerrit.fd.io/r/gitweb?p=csit.git;a=tree;f=PyPI/MLRsearch;hb=refs/heads/master
-.. _IETF draft: https://tools.ietf.org/html/draft-ietf-bmwg-mlrsearch-00
+.. _IETF draft: https://tools.ietf.org/html/draft-ietf-bmwg-mlrsearch-04