Changes between Initial Version and Version 1 of UserGuide/BandwidthTweaking


Ignore:
Timestamp:
02/19/2010 02:59:13 AM (14 years ago)
Author:
pi3832
Comment:

I could be completely wrong. But I thought I would put something together based on my current understanding.

Legend:

Unmodified
Added
Removed
Modified
  • UserGuide/BandwidthTweaking

    v1 v1  
     1== Bandwidth Tweaking == 
     2 
     3The settings below are suggestions.  They are intended to give users more guidance than the settings recommended in the FAQ, but are by no means definitive. 
     4 
     5'''Don't be greedy.'''  There's more content out there than you can ever consume.  You don't need to download it all '''''right now!!'''''  By allocating your bandwidth properly you can maintain a solid, thick stream of data that fills up your hard-drive 10-times faster than you'd expect. 
     6 
     7Maximizing your peak download rates can be fun and all, but it doesn't necessarily get you the best over-all performance. 
     8 
     9=== Determine Internet connection speed === 
     10 
     11Go to [http://speedtest.net/ SpeedTest.net] to test your connection.  It's best to do this with all other applications, including Deluge, closed.  You also should run the test a few times, hours or days apart, to make sure your initial results were accurate. 
     12 
     13!SpeedTest.net will give you results in Mbps (Mb/s).  Multiply those results by 122 to get your connection speed in KiB/s, which are the units used in the Preferences->Bandwidth window.  (At least they are for v1.2.0 for Linux.) 
     14 
     15=== Preferences -> Bandwidth === 
     16 
     17==== Global Settings ==== 
     18 
     19|| Maximum Connections || -1 || (Let the per-torrent settings limit the number of connections.) || 
     20[[BR]] 
     21|| For upload speeds of: || < 50 KiB/s || 50 - 150 KiB/s || 150 - 250 KiB/s || > 250 KiB/s || 
     22|| Maximum Upload Slots || 2 * || 4 * || 7 * || 10 * || 
     23[[BR]] 
     24|| Maximum Download Speed || 80 - 95% of tested download speed || (Be careful of setting this too high--it can strangle your browsing, media streaming, etc. and other people who use your connection.) 
     25|| Maximum Upload Speed || 80% of tested upload speed || (Upload speed is the real limiter for most things internet--beware guarantees of huge download speeds that don't mention upload speeds.) || 
     26|| Maximum Half-Open Connections: || 50 - 100 || (Unless you've got a Windows install with limited half-open connections.  Then this should be 50 - 80% of that limit.) || 
     27 
     28 Everyone got all excited when it was discovered that Microsoft had, at one point, limited half-open connections in XP and Vista to small numbers.  As a plethora of hacks came out to remove this limit, somehow "half-open connections" became the scape-goat for slow download speeds.  As people hacked around the limit, and as Microsoft removed it in later release, it seemed that, if limited half-open connections is bad, then lots and lots of them must be good.  Not really, though.   
     29 
     30 Half-open connections should either resolve to fully-opened connections or be timed-out.  So, you really don't need to allow that many of them to be hanging around. 
     31 
     32|| Maximum Connection Attempts per Second || 20 || (You don't need to fill all your allowed connections in the first second, do you?) || 
     33 
     34 
     35==== Per Torrent Settings ==== 
     36 
     37|| For upload speeds of: || < 50 KiB/s || 50 - 150 KiB/s || 150 - 250 KiB/s || > 250 KiB/s || 
     38|| Maximum Connections || 30 || 50 || 100 || 200 || 
     39|| Maximum Upload Slots || 2 || 4 || 7 || 10+ || 
     40 
     41While a basic premise of bit-torrenting is the "swarm" of peers, with more being better, you, as a single client, can spread yourself too thin.  You're more helpful to the swarm by feeding a limited number of peers with a steady, healthy stream of data, than you are by trickling out data to a huge number of peers. 
     42 
     43You should limit your upload slots based on your upload speeds to make sure that each connected peer is getting a reasonable amount of bandwidth. 
     44 
     45You should limit your number of connections, because it does take resources to keep track of each connection, and why track connections that are giving you, relatively, minor speeds?  One peer feeding you 5 KiB/s is worth 50 peers flickering at 0.1 KiB/s.  (Why are there so many 0.1 KiB/s feeds out there?  Perhaps because too many people aren't setting their "upload slots" correctly?) 
     46 
     47You also don't want/need to be keeping track of a large number of peers waiting to leech when you have limit upload slots. 
     48 
     49|| Maximum Download Speed || -1 || (Let the per-torrent speed by limited by the global settings.) || 
     50|| Maximum Upload Speed || -1 || (Let the per-torrent speed by limited by the global settings.) || 
     51 
     52You can, by setting the per-torrent maximums to less than the global maximums, prevent a single torrent from using up all the allocated bandwidth and forcing all of the other (auto-managed) torrents to pause.  But there is no advantage to increasing your number of active torrents.  Indeed, by letting the torrent with the best speeds dominate, it will finish as quickly as possible and then stop competing at all for resources with the other torrents in the queue. 
     53 
     54 
     55 
     56=== Preferences -> Queue === 
     57 
     58==== Active Torrents ==== 
     59 
     60|| For upload speeds of: || < 50 KiB/s || 50 - 150 KiB/s || 150 - 250 KiB/s || > 250 KiB/s || 
     61|| Total active || || || || || 
     62|| Total active downloading || || || || || 
     63|| Total active seeding: || || || || || 
     64 
     65All downloading torrents are automatically also seeding.  This means that the ''actual'' number of maximum seeding torrents the number of "active seeding" '''plus''' "active downloading." 
     66 
     67So, if you have limited bandwidth, you can end up strangling your active seeds by having too many active downloads.