Fstress is a benchmark that is used to test NFS server performance (average 
latency for a mixed number of client operations at different load levels). 
The source code here were obtainded from Duke, with several small changes.
Many thanks to its authors, Darrell Anderson and Jeff Chase, for their permissions
for us to provide the benchmark here. For more detailed documents, please
go to the fstress webpage at Duke: http://www.cs.duke.edu/ari/fstress/.

1. BUILD
	export FSTRESS_HOME=/fstress   // assume the path of fstress code is /fstress
	cd $FSTRESS_HOME; make

2. CONFIGURE
	Edit ~/.bashrc file, adding the following line:
		export FSTRESS_HOME=/fstress
	Edir ~/.ssh/environment, adding the following line:
		FSTRESS_HOME=/fstress
	Fstress allows using multiple clients during a test. A master machine starts executions 
	on the specified client machines. To do this, the master needs to be able to execute 
	command remotely. The default remote shell fstress uses is rsh. To use ssh instead, 
	use option -ssh when starting fstress. Also, use ssh-keygen to setup login without passwd:
		ssh-keygen -t dsa   // Just enter for any questions ssh-keygen asks, execute this on each client
		scp ~/.ssh/id_dsa.pub master-host:/home-dir/.ssh/client-host.pub   // scp the generated id_dsa.pub to the master
		cat /home-dir/.ssh/client-host.pub >> /home-dir/.ssh/authorized_keys2   // add to authorized_keys2
	Make sure the used client machines have access (can mount and write) on NFS servers

3. RUN
	Below are some examples of using fstress to measure NFS server response time. 
	For a complete list of options, please refer to the manual.pdf included in this directory.
	/* run fstress on a single NFS client (localhost) with NFS server server-name */
		/fstress/bin/fstress.csh -ssh -clients localhost -server server-name:/nfs-test
	/* run fstress with starting load 100 (200 ops/s), ending load 1000, increasing 100 per run */
		/fstress/bin/fstress.csh -low 1000 -high 1000 -incr 100 -ssh -clients localhost -server server-name:/datasrc
	/* run fstress with warmup time as 500s (default 300s), run time 500s (default 300s) */
		/fstress/bin/fstress.csh -warmup 500 -run 500 -ssh -clients localhost -server server-name:/datasrc
	/* specify maximum latency */
		/fstress/bin/fstress.csh -maxlat 1000 -ssh -clients localhost -server server-name:/datasrc
	/* have fstress report stats every minute and run it 4 hours at load 1000, loop is the number of repeats. i.e., total-time/run-time */
		/fstress/bin/fstress.csh -low 1000 -high 1000 -maxlat 1000 -run 60 -loop 240 -ssh -clients localhost -server server-name:/nfs-test

4. VIEW RESULTS 
	Run fstress_multistich.csh to view the results. output is the directory that contains the running results
	/fstress/bin/fstress_multistich.csh output

5. Evaluate ZUMASTOR PERFORMANCE
	This directory includes an example script, run.sh, for comparing fstress performance with and without zumastor.
	Have a look at the script and change the settings in it according to your nfs server setup. Then execute the
	script to run the test:
		./run.sh
	After the tests finish, use another included script to view the results:
		./generate_all_gnuplot.sh fstress-output | gnuplot
	where fstress-output is the directory that contains the generated benchmark results.