[R] scaling to multiple data files

Jason Edgecombe jason at rampaginggeek.com
Tue Jan 11 21:36:23 CET 2011

That's correct, those users have been logged in or had processes running 
on this machine for four days. The machines in question are time-sharing 
Linux servers for college students and professors to use. multi-day jobs 
are common.

The "last" command does what you suggest, but it doesn't capture 
processes left running in the background when a user logs out.  This 
data is simpler and collected across Windows and Linux hosts. Sessions 
are somewhat ambiguous. We just care who is running processes on a 
machine at a certain time. We also collect the process name of processes 
that each user is running so that we can gauge how often applications 
are being used, and by whom. For this analysis, I'm not worried about 
which processes are running, only unique users per day. i have almost 
four years of historical data for some machines in this format.

We have multiple tools written in different languages that parse this 
data. I'm writing one that does better graphing.

On 01/11/2011 11:39 AM, jim holtman wrote:
> I am not sure exactly what your data represents.  For example, from
> looking at the data it appears that user1 and user2 have been logged
> on for about 4 days; is that what the data is saying?  If you are
> keeping track of users, why not write out a file that has the
> start/end time for each user's session.  The first time you see them,
> put an entry in a table and as soon as they don't show up in your
> sample, write out a record for them.  With that information is it easy
> to create a report of the number of unique people over time.
> On Tue, Jan 11, 2011 at 10:47 AM, Jason Edgecombe
> <jason at rampaginggeek.com>  wrote:
>> Hello,
>> I have logging information for multiple machines, which I am trying to
>> summarize and graph. So far, I process each host individually, but I would
>> like to summarize the user count across multiple hosts. I want to answer the
>> question "how many unique users logged in on a certain day across a group of
>> machines"?
>> I'm not quite sure how to scale the data frame and analysis to summarize
>> multiple hosts, though. I'm still getting a feel for using R.
>> Here is a snippet of data for one host. the user_count column is generated
>> from the users column using my custom function "usercount()". the samples
>> are taken roughly once per minute and only unique samples are recorded.
>> (i.e. use na.locf() to uncompress the data). Samples may occur twice in the
>> same minute and are rarely aligned on the same time.
>> Here is the original data before I turn t into a zoo series and run
>> na.locf() over it so I can aggregate a single host by day. I'm open to a
>> better way.
>>> foo
>>                   users            datetime user_count
>> 1         user1&  user2 2007-03-29 19:16:30          2
>> 2         user1&  user2 2007-03-31 00:04:46          2
>> 3         user1&  user2 2007-04-02 11:49:20          2
>> 4         user1&  user2 2007-04-02 12:02:04          2
>> 5         user1&  user2 2007-04-02 12:44:02          2
>> 6 user1&  user2&  user3 2007-04-02 16:34:05          3
>>> dput(foo)
>> structure(list(users = c("user1&  user2", "user1&  user2", "user1&  user2",
>> "user1&  user2", "user1&  user2", "user1&  user2&  user3"), datetime =
>> structure(c(1175210190,
>> 1175313886, 1175528960, 1175529724, 1175532242, 1175546045), class =
>> c("POSIXt",
>> "POSIXct"), tzone = "US/Eastern"), user_count = c(2, 2, 2, 2,
>> 2, 3)), .Names = c("users", "datetime", "user_count"), row.names = c(NA,
>> 6L), class = "data.frame")
>> Thanks,
>> Jason
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.

More information about the R-help mailing list