Replace reliability optimizations with built-in profile Optimization.ReliabilityProtocol.Common

Signed-off-by: Andrea Sorbini <[email protected]>

Author: asorbini

README.md

@@ -210,7 +210,7 @@ its defaults.
 
 ### RMW_CONNEXT_DISABLE_RELIABILITY_OPTIMIZATIONS
 
-By default, `rmw_connextdds` will modify the QoS of each realiable DataWriter
+By default, `rmw_connextdds` will modify the QoS of each reliable DataWriter
 and DataReader to improve the responsiveness of the RTPS [reliability protocol](https://community.rti.com/static/documentation/connext-dds/6.0.1/doc/manuals/connext_dds/html_files/RTI_ConnextDDS_CoreLibraries_UsersManual/Content/UsersManual/Using_QosPolicies_to_Tune_the_Reliable_P.htm?tocpath=Part%203%3A%20Advanced%20Concepts%7C11.%20Reliable%20Communications%7C11.3%20Using%20QosPolicies%20to%20Tune%20the%20Reliable%20Protocol%7C_____0#reliable_1394042328_776265).
 
 For example, the ["heartbeat period"](https://community.rti.com/static/documentation/connext-dds/6.0.1/doc/manuals/connext_dds/html_files/RTI_ConnextDDS_CoreLibraries_UsersManual/Content/UsersManual/Controlling_Heartbeats_and_Retries.htm#reliable_1394042328_785637)

rmw_connextdds_common/include/rmw_connextdds/static_config.hpp

@@ -238,51 +238,6 @@
 #define RMW_CONNEXT_DEFAULT_RELIABILITY_OPTIMIZATIONS     1
 #endif /* RMW_CONNEXT_DEFAULT_RELIABILITY_OPTIMIZATIONS */
 
-/******************************************************************************
- * Regular hearbeat period used by any reliable RTPS Writer.
- * This is an initializer for an instance of type DDS_Duration_t.
- ******************************************************************************/
-#ifndef RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD
-#define RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD       {0, 100000000}   /* 100ms */
-#endif /* RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD */
-
-/******************************************************************************
- * Fast hearbeat period used by any reliable RTPS Writer to allow
- * late joiners and out of sync readers to catch up.
- * This is an initializer for an instance of type DDS_Duration_t.
- ******************************************************************************/
-#ifndef RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD_FAST
-#define RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD_FAST  {0, 20000000}   /* 20ms */
-#endif /* RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD_FAST */
-
-/******************************************************************************
- * When a DataWriter receives a request for missing DDS samples from a
- * DataReader and responds by resending the requested DDS samples, it will
- * ignore additional requests for the same DDS samples during the time period
- * max_nack_response_delay. We decrease this to be less than the HB period.
- ******************************************************************************/
-#ifndef RMW_CONNEXT_DEFAULT_MAX_NACK_RESPONSE_DELAY
-#define RMW_CONNEXT_DEFAULT_MAX_NACK_RESPONSE_DELAY  {0, 10000000}   /* 10ms */
-#endif /* RMW_CONNEXT_DEFAULT_MAX_NACK_RESPONSE_DELAY */
-
-/******************************************************************************
- * Maximum number of periodic heartbeats gone unanswered after which a
- * DataWriter will consider a DataReader as inactive.
- *
- ******************************************************************************/
-#ifndef RMW_CONNEXT_DEFAULT_MAX_HEARTBEATS
-#define RMW_CONNEXT_DEFAULT_MAX_HEARTBEATS    (10 * 60) /* 1m @ 10hz */
-#endif /* RMW_CONNEXT_DEFAULT_MAX_HEARTBEATS */
-
-/******************************************************************************
- * When a reliable reader receives a heartbeat from a remote writer and finds
- * out that it needs to send back an ACK/NACK message, the reader can choose to
- * delay a while. We set this delay to be compatible with the HB period.
- ******************************************************************************/
-#ifndef RMW_CONNEXT_DEFAULT_MAX_HEARTBEAT_RESPONSE_DELAY
-#define RMW_CONNEXT_DEFAULT_MAX_HEARTBEAT_RESPONSE_DELAY  {0, 10000000} /* 10ms */
-#endif /* RMW_CONNEXT_DEFAULT_MAX_HEARTBEAT_RESPONSE_DELAY */
-
 /******************************************************************************
  * Automatically tune DataWriterQos to better handle reliable "large data".
  ******************************************************************************/

rmw_connextdds_common/src/ndds/dds_api_ndds.cpp

@@ -504,21 +504,26 @@ rmw_connextdds_get_datawriter_qos(
 #if RMW_CONNEXT_DEFAULT_RELIABILITY_OPTIMIZATIONS
   // The default settings for the RTPS reliability protocol are not very
   // responsive, and they cause some unit tests to fail. These optimizations
-  // increase the Heartbeat period from 3s (default) to 100ms. Other vendors
-  // seem to be in a similar range. We also lower the period for "late joiners"
-  // and the "fast" period (used to speed up recovery of readers that reached
-  // the "high watermark" of unacked samples -- 1 by default) to 20ms.
+  // have been derived from profile `Optimization.ReliabilityProtocol.Common`
+  // available in Connext 6+. `Generic.StrictReliable` is the equivalent
+  // profile in 5.3.1. Changes are limited to `DDS_RtpsReliableWriterProtocol_t`.
   if (ctx->optimize_reliability) {
-    qos->protocol.rtps_reliable_writer.heartbeat_period =
-      RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD;
-    qos->protocol.rtps_reliable_writer.late_joiner_heartbeat_period =
-      RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD_FAST;
-    qos->protocol.rtps_reliable_writer.fast_heartbeat_period =
-      RMW_CONNEXT_DEFAULT_HEARTBEAT_PERIOD_FAST;
-    qos->protocol.rtps_reliable_writer.max_heartbeat_retries =
-      RMW_CONNEXT_DEFAULT_MAX_HEARTBEATS;
-    qos->protocol.rtps_reliable_writer.max_nack_response_delay =
-      RMW_CONNEXT_DEFAULT_MAX_NACK_RESPONSE_DELAY;
+    // All write() calls will block (for at most max_blocking_time) once the send_window
+    // is filled with samples that haven't yet been acknowledged by all active readers.
+    qos->protocol.rtps_reliable_writer.min_send_window_size = 40;
+    qos->protocol.rtps_reliable_writer.max_send_window_size = 40;  // fixed size window
+    qos->protocol.rtps_reliable_writer.heartbeats_per_max_samples = 10;  // 1 every 4
+    qos->protocol.rtps_reliable_writer.heartbeat_period = {0, 200000000};  // 200ms
+    qos->protocol.rtps_reliable_writer.late_joiner_heartbeat_period = {0, 20000000};  // 20ms
+    qos->protocol.rtps_reliable_writer.fast_heartbeat_period = {0, 20000000};  // 20ms
+    qos->protocol.rtps_reliable_writer.max_heartbeat_retries = 500;  // 10s @ 50hz
+    // Force the writer to reply immediately to ACKNACK's received from a writer.
+    qos->protocol.rtps_reliable_writer.max_nack_response_delay = DDS_DURATION_ZERO;
+    // When the number of unack'd samples reaches the high_watermark the fast_heartbeat_period
+    // is used. When the number dips below the low_watermark, the heartbeat_period is used.
+    // These numbers are tied to the send_window size.
+    qos->protocol.rtps_reliable_writer.high_watermark = 25;
+    qos->protocol.rtps_reliable_writer.low_watermark = 10;
   }
 #endif /* RMW_CONNEXT_DEFAULT_RELIABILITY_OPTIMIZATIONS */
 
@@ -612,10 +617,10 @@ rmw_connextdds_get_datareader_qos(
   // The default settings for the RTPS reliability protocol are not very
   // responsive, and they cause some unit tests to fail. These optimizations
   // are dual to those applied in rmw_connextdds_get_datawriter_qos().
+  // Changes are limited to `DDS_RtpsReliableReaderProtocol_t`.
   if (ctx->optimize_reliability) {
     qos->protocol.rtps_reliable_reader.min_heartbeat_response_delay = DDS_DURATION_ZERO;
-    qos->protocol.rtps_reliable_reader.max_heartbeat_response_delay =
-      RMW_CONNEXT_DEFAULT_MAX_HEARTBEAT_RESPONSE_DELAY;
+    qos->protocol.rtps_reliable_reader.max_heartbeat_response_delay = DDS_DURATION_ZERO;
   }
 #endif /* RMW_CONNEXT_DEFAULT_RELIABILITY_OPTIMIZATIONS */