Clustering updates
Yesterday (wednesday) and a few hours this morning, I toyed around with some code generating live statistics on our new [“brand new clustering solution”:http://printf.dk/work/clustering.html]. It generates some graphs, but since we’re using DR(Direct Routing) it’s impossible to meassure the outgoing traffic as well as the incoming. The graphs and charts are pretty interesting though.
What’s really interesting is the fact that the piechart of connection distribution is exactly 33.33% divided to three servers. That’s nothing short of perfect loadbalancing. Also, the bytes-per-packet averages pretty well on each server too. Each server has had roughly even amounts of incoming traffic, this being caused by the “wlc” algorhitm on the LVS master. Enough talk, here’s a screenshot: !http://printf.dk/img/lvsstats-thumb.png!:http://printf.dk/img/lvsstats.png