From 12f9915d1c8b0164517ee7206c1ba644b55a8f5a Mon Sep 17 00:00:00 2001
From: Richard Hellwig <richard.hellwig@tu-dortmund.de>
Date: Wed, 12 Jun 2013 15:51:07 +0200
Subject: [PATCH] core/efw: send back results earlier

The client sends results back earlier (i.e., before all jobs are
done) if the client response time (CLIENT_JOB_REQUEST_SEC) is
exceeded. This makes sure that extraordinarily long-running
experiments get reported back before, e.g., the LIDO job timeout
kills the Fail* instance.

Change-Id: I3ada0360ec54b63f80a7008570ca514449720220
---
 src/core/efw/JobClient.cc  | 17 +++++++++++++++--
 src/core/efw/JobClient.hpp |  2 ++
 2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/src/core/efw/JobClient.cc b/src/core/efw/JobClient.cc
index 2df2171b..153c863d 100644
--- a/src/core/efw/JobClient.cc
+++ b/src/core/efw/JobClient.cc
@@ -182,12 +182,23 @@ bool JobClient::sendResult(ExperimentData& result)
 	m_results.push_back( temp_exp );
 
 	if (m_parameters.size() != 0) {
+		//If job request time is over send back all existing results
+		if (CLIENT_JOB_REQUEST_SEC < (double)m_job_runtime) {
+			m_job_runtime_total += (double) m_job_runtime;
+			m_job_runtime.reset();
+			m_job_runtime.startTimer();
+			m_job_total += m_results.size();
+			sendResultsToServer();
+		}
+
 		//If there are more jobs for the experiment store result
 		return true;
 	} else {
 		//Stop time measurement and calculate new throughput
 		m_job_runtime.stopTimer();
-		m_job_throughput = 0.5 * m_job_throughput + 0.5*(CLIENT_JOB_REQUEST_SEC/((double)m_job_runtime/m_results.size()));
+		m_job_runtime_total += (double) m_job_runtime;
+		m_job_total += m_results.size();
+		m_job_throughput = 0.5 * m_job_throughput + 0.5*(CLIENT_JOB_REQUEST_SEC/(m_job_runtime_total/m_job_total));
 
 		if (m_job_throughput > CLIENT_JOB_LIMIT) {
 			m_job_throughput = CLIENT_JOB_LIMIT;
@@ -195,8 +206,10 @@ bool JobClient::sendResult(ExperimentData& result)
 			m_job_throughput = 1;
 		}
 
-		//Reset timer for new time measurement
+		//Timer/Counter cleanup
 		m_job_runtime.reset();
+		m_job_runtime_total = 0;
+		m_job_total = 0;
 
 		return sendResultsToServer();
 	}
diff --git a/src/core/efw/JobClient.hpp b/src/core/efw/JobClient.hpp
index 8ff9d174..de4153c5 100644
--- a/src/core/efw/JobClient.hpp
+++ b/src/core/efw/JobClient.hpp
@@ -31,7 +31,9 @@ private:
 	uint64_t m_server_runid;
 
 	WallclockTimer m_job_runtime;
+	double m_job_runtime_total;
 	int m_job_throughput;
+	int m_job_total;
 	std::deque<ExperimentData*> m_parameters;
 	std::deque<ExperimentData*> m_results;