Off Topic 
Important Traffic results for Server Hosters
1 
Important Traffic results for Server Hosters
Offline
Article: http://tgvserver.info/analysisHourTraffic.pdf
Note - while the study itself is around 70 pages, the analysis and conclusion span the first 9 pages only. Appendix A also gives marginal analysis of volume traffic at the high and low traffic time as well as a Day Of Week analysis (The highest traffic usually occurs on Saturdays). Source-code for extracting bandwidth data is found in Appendix B
Abstract:
It’s often difficult for game server providers to allocate enough resources for a multitude of servers that they are required to host. A workaround to this limitation of virtual resources is by automatically scaling the resource limitation for each node of the server in such a way that an idle server would require less resource than a busy server. However, a problematic aspect of this approach is that it does not look into the fundamental problem of not having enough resources to begin with to process the incoming and outgoing data. If for any reason a sudden traffic spike hits a significant number of servers, including malicious events such as a DDoS attack concentrated on an 8-bit IP block, and then the servers will invariably fail leading to potential client loss. It’s thus invaluable for server providers to have an accurate estimate of when the margins of traffic will occur and to use that knowledge to their advantage.
This paper will analyze any possible association between the volume of traffic (Upload Traffic) to a game server and the hour of the day. The data will be sampled using an automated data collection tool (Source Code provided in Appendix) which will extract the traffic data from a server-generated report and create a SQL query which will be transformed into an xlsx spreadsheet. The traffic data will be sampled from February 2009 up to May 2009 (February 1st through April 30th). However, due to the fact that there are inevitable downtimes due to either 1) downtime from the Internet Provider 2) downtime from Server Collocation Cluster or 3) Emergency Reboot, there will be gaps within the data collected. The effects of these gaps are subject to analysis, but for the sake of this observation, they are insignificant.