Tor seems slower than my normal direct internet connection, perhaps because of high traffic from users or low bandwidth.
Can anyone explain the speed differences to me?
Thanks so much. Have a nice day.
External research performed December 2010 to January 2011 found that "The average ratio of HTTP request durations (Tor/direct) [from around the world] is 4.1."
"Another observation is that the four nodes located in Germany, UK, and USA always have the fastest Tor connection compared to the other nodes.. more than 50% of all Tor nodes [were] located in these countries [at the time]." Today, Germany, France, and Netherlands surpass the United States.
Also, "During the timeframe of our experiments, the median value for downloading of 50 KB was reported [by Tor Metrics] as between 3 and 4 s, and for downloading of 1 MB between 9 and 15 s. Our results of downloading 50 KB were between 3 and 11 s (median), and for the download of 1 MB between 12 and 26 s (median). We see them as comparable."
Note that today, Tor Metrics reports a median value for downloading 1 MB between 1.00 and 1.25 seconds. This is an improvement by a factor of 10 from the time of the research cited!
As for bandwidth, Tor Metrics reports that approximately 50% of advertised network bandwidth is currently utilized.
Tor is somewhat slow by design. Bouncing your traffic between three nodes before sending it out to the wider internet (6 nodes after the response is received) adds a significant amount of latency to the round trip time.
To increase the bandwidth alloted to a node you run, look at the following lines in your
torrc file (with some large example values):
RelayBandwidthRate 300 KB RelayBandwidthBurst 350 KB MaxAdvertisedBandwidth 100 KB
Manually picking three relays and using them removes almost all of the benefits Tor provides. Now an attacker can simply focus on those three nodes. Persuading your friends to turn over access to their nodes, or just stealing their private keys (or performing a DOS attack, or some other form of attack) is probably going to be easier for the attacker than if you're using three random nodes as is intended.
This 2009 Tor Project blog post on the subject might interest you.