question
91 views
Performance Testing
Hello,
Does the performance testing for m2 need to be for both latency and throughput? Or can it just be for one of them like in m1?
Thanks!
the students' answer,
where students collectively construct a single answerthe instructors' answer,
where instructors collectively construct a single answerYou need to measure both the latency and the throughput of your system.
followup discussionsfor lingering questions and comments
Anonymous Atom
Is it okay if we just use a portion of the enron dataset given the size of the file?
Gengrui Zhang
yes
Anonymous Helix
Do these performance tests also need to measure different put/get splits? Ie. Test 20/80, 50/50, 80/20 splits for each case of 1, 5, ,20, ... clients and 1,5,10,50 ... servers and cache configurations for a total of = n1 put/get splits x n2 # clients x n3 # servers x n4 # server configurations?
In addition, how many combinations/total tests should we have to populate our report?
Thanks in advance!
Gengrui Zhang
Generally yes. I know this is a lot when you multiply them, so I would suggest you have one caching strategy, three splits, two configs for clients (maybe 1 and 20?) and three server configs (1, 5, and 20). Write a script and let it run automatically, each run for 30s (if disks can hold) and calculate the average.