Final presentation

Download Report

Transcript Final presentation

NETWORK
TRAFFIC
Paul German, Jeffrey Klow, and
Emily Andrulis
THE IDEA
 Ben’s proposal
 Examine Cornell’s Network Traffic
 How much do we use? When do we use it?
 What information can we glean?
OUR MAIN QUESTIONS
 What does an average day at Cornell look like in
regards to network traffic?
 Assuming the pattern holds, at what point should we
consider getting more bandwidth because we will be
frequently coming close to our maximum allotted?
GETTING THE DATA
 The Tims in Network Services
 Log data files for primary and secondary internet
provider, and internal network traffic
 Log files include upload and download averages and
maximums
 Decreasing time resolution between lines
 Solution: Collect data for 1 week around same time
each day
DATA CLEANING
 Create scripts in R
 log file -> data frames in R
 Update already made data frames with new log data
 Add different time variables
 UNIX -> CST, date, time, weekday, decimal time
 Add % of bandwidth variables
 Helper functions
 getSelectedIndices
 modifyDataResolution
TELLING THE STORY
 Use static, animated, and interactive graphs to display
data
 Go back to our focus questions:
 Average day at Cornell?
 Frequency of reaching 85% bandwidth?
 What does the future usage look like?
EXPLAINING THREE TYPES
 Log files from primary internet provider, secondary
internet provider, and internal network traffic
 Cap differences: 300 Mb/sec vs. 100 Mb/sec
 Internal weird
AVERAGE USAGE AT CORNELL
 Static -> Interactive
AVERAGE USAGE SECONDARY
AVERAGES THROUGH ANIMATION
 Day of Week compared to Average Day
AVERAGE LAST WEEK
 Average Day compared to Days Last Week
AVERAGES SINCE NOVEMBER
 Average Day compared to all days back to November
AVERAGE BLOCK USAGE
 Showing Usage over Block 4
BLOCK 4 SECONDARY
 Block 4 Usage on Secondary Provider (Note: peaks)
WHERE ARE WE HEADING?
FUTURE APPLICATIONS
 Give code to the Tims
 Documented and split up by task
 Interactive graphs with new data
 Easily replicable
 Raise awareness about usage in terms of averages and
when we’re nearing the cap