Jump to content


Photo

Volunteers Needed for Research Data Collection


  • Please log in to reply
8 replies to this topic

#1 Roggz

Roggz

    Glass Joe

  • Members
  • 6 posts

Posted 16 September 2010 - 06:39 PM

I am a graduate student working on a open, programmable, scalable, secure, and extensible virtual environment called Meru.

One aspect of this virtual environment is application-level messaging. Users can create objects in our world, and these objects can communicate with each other. As part of our research, we'd like to collect some data from a real virtual world like World of Warcraft.

This is where you come in. If you are willing to help us out, I have written an AddOn for WoW called AppMsgLogger. This AddOn collects three pieces of information:

  • A timestamp every time you log in or log out
  • For every addon message you receive (the CHAT_MSG_ADDON event), a timestamp, the zone you are in, the channel it was received on (PARTY, GUILD, RAID, etc), and the length of the message in bytes
  • A list of the AddOns you are running
No personal or identifying information is collected.

An explanation of why we want to collect each piece of information:
  • is just so I can know how long the addon was running for.
  • is what we're really interested in. It allows us to model what
    application messaging looks like in a virtual world. We can calculate
    things like how many bytes/second the average player receives in
    application messages, what application message traffic looks like
    based on what type of activity the user is doing (pvp, pve, city,
    etc), or a breakdown of which channels applications are using for
    their messages.
  • is just so we can tell which messages are actually being received.
    If anyone in your guild/party/raid has an addon that sends a message,
    you receive it regardless of whether you're even running that addon
    which receives the message.

So you know I'm legit,
Here's me: Jeff Terrace
Here's the project: SING: Meru Virtual World Architecture
Here's a paper that we previously published: http://www.cs.prince...rlds-ieee09.pdf

I'm also happy to release the dataset into the public domain, as I think it has value beyond what we're using it for. If I get enough useful data, I will release it.

If you would like to help us out by collecting some data, please visit:
WoW Data Collection - AppMsgLogger

And before I'm asked:
  • No, this is not spam
  • No, I'm not trying to steal your WoW account
  • Yes, I made sure to get permission from an admin (Boethius) before posting

Thank you!

Jeff

#2 Roggz

Roggz

    Glass Joe

  • Members
  • 6 posts

Posted 22 September 2010 - 04:58 PM

In case anyone is interested, here are some interesting things I've found so far.

  • This graph shows a histogram of message sizes. It shows that the vast majority of messages sent are extremely small, with a cluster right around the maximum message size of 256 bytes: Posted Image
  • This graph plots how often each addon send messages. It shows that the distribution is exponential (log-log plot): that is, the most popular addon sends an extremely large amount of messages, while there is a heavy tail of addons that send message very infrequently: Posted Image
  • To support our hypothesis that traffic is bursty, I plotted the average incoming byte rate in 1-minute intervals over 24 hours of play time. The graph shows that traffic is indeed bursty, with some periods showing up to 14Kbit/s while others are almost 0:Posted Image
  • We also wanted to see what percentage of traffic is wasted. Each data file has the list of addons the user was running, so by matching message prefix with addon names, I can determine what percentage of the messages received were wasteful (in that the intended recipient addon was not installed by the user). This has been highly variables so far. Out of the users who have sent me data, the percentages are: 58.86%, 28.19%, 15.87%, 12.63%, 9.52%, 86.08%, 83.92%, 17.03%. This is where more data points would help significantly. If you have data to send me, please do!


#3 Kevinally

Kevinally

    Bald Bull

  • Members
  • 1,045 posts

Posted 22 September 2010 - 08:16 PM

In regards to point 3, you said you plotted over a 24 hour play period, but I only see the chart going to 1600. Was 1600-2359 clipped off, or am I misreading the legend?

#4 Kadeth

Kadeth

    Glass Joe

  • Members
  • 1 posts

Posted 22 September 2010 - 08:43 PM

There is only 1440 minutes in a day.

#5 Kevinally

Kevinally

    Bald Bull

  • Members
  • 1,045 posts

Posted 22 September 2010 - 09:06 PM

Ah, so you're going by raw minutes. I was assuming it was military time - Thanks for the clarification.

In that case, do you have any hypotheses for these peaky times? My initial thought was popular raid times, but the 0 to ~180 chunk seemed a bit strange to me. Are you willing to say how large your data set is at this point?

#6 Roggz

Roggz

    Glass Joe

  • Members
  • 6 posts

Posted 22 September 2010 - 10:59 PM

Ah, so you're going by raw minutes. I was assuming it was military time - Thanks for the clarification.

In that case, do you have any hypotheses for these peaky times? My initial thought was popular raid times, but the 0 to ~180 chunk seemed a bit strange to me. Are you willing to say how large your data set is at this point?


Ah, I can see how that was misleading. Yes, those are absolute minutes, not military time. Also, this is 24 hours of play time for a single user with multiple sessions, strung together. That's why it probably looks a little odd. Here's the same graph (actually corrected, as the previous one had an error in it) with altering colors when the user changed zones. I also included the name of the zone in the center of the colored band when the total duration in that zone was > 10 minutes.

Posted Image

#7 Kevinally

Kevinally

    Bald Bull

  • Members
  • 1,045 posts

Posted 23 September 2010 - 01:46 AM

I see. Once again, I misread the legend in assuming that it covered the average daily use of multiple people. Also, your coloring answers another question, but just to be clear - this is the usage graph of only one player, correct? Would it be possible to get an average use graph, laid out over a full 24 hour day? I don't know how possible that would be for you, but I could see the data set as useful.

#8 Roggz

Roggz

    Glass Joe

  • Members
  • 6 posts

Posted 23 September 2010 - 02:59 AM

but just to be clear - this is the usage graph of only one player, correct?

Yep

Would it be possible to get an average use graph, laid out over a full 24 hour day?


Yes, but it wouldn't be valid. I don't have enough data for it to be meaningful.

#9 Shugoshin

Shugoshin

    Glass Joe

  • Members
  • 6 posts

Posted 13 October 2010 - 03:30 PM

I'm always happy to see projects like these. I've taken a few research classes for Criminal Justice and found a few aspects very interesting.

I do have an observation. You seem to have a lot of data pointing to raid chat, guild chat, and activity in Icecrown. Have you been reaching out to the casual community as well, perhaps posting on other sites? Or is it an intent for you to focus on the more active, end-game community? In the past, the majority of the WoW player base has not been involved in raiding. I can't find a good current source for you on this, but it may be worth looking into. Otherwise, perhaps in any conclusions you make you may want to point out where your majority of volunteers may have come from and how it could have affected your data.

Just my 2 cents. Good luck! I'll be sure to get involved myself.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users