I QOS filtered an entire subnet to use half my max upload, and low
priority. my other subnet is configured to have up to 100% and high
priority.
wan0 1 4*full/10 full 1
default,tos-minimize-delay
wan0 2 4*full/10 full 2
tos-maximize-throughput
wan0 3 2*full/10 5*full/10 3
traffic ruled into mark 3 in tcrules includes a dedicated bittorrent
server. for some reason, it still manages to slow category 1 and 2 to
a crawl :( what might I be doing wrong ?
--
(\_/) This is Bunny. Copy and paste Bunny
(=''.''=) into your signature to help him gain
(")_(") world domination.
------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
Christ Schlacta wrote:> I QOS filtered an entire subnet to use half my max upload, and low > priority. my other subnet is configured to have up to 100% and high > priority. > > wan0 1 4*full/10 full 1 > default,tos-minimize-delay > wan0 2 4*full/10 full 2 > tos-maximize-throughput > wan0 3 2*full/10 5*full/10 3 > > > traffic ruled into mark 3 in tcrules includes a dedicated bittorrent > server. for some reason, it still manages to slow category 1 and 2 to > a crawl :( what might I be doing wrong ?These problems are often caused by using bandwidth settings that are not realistic (e.g., what your ISP says your bandwidth is rather than what it actually is). The settings in /etc/shorewall/tcdevices should always be very conservative. Without details, we really can''t advise you further. Please see http://www.shorewall.net/support.htm#Guidelines. -Tom -- Tom Eastep \ When I die, I want to go like my Grandfather who Shoreline, \ died peacefully in his sleep. Not screaming like Washington, USA \ all of the passengers in his car http://shorewall.net \________________________________________________ ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects
http://bitch.aarcane.info/shorewall_dump also, I've acquired the numbers in tcdevices from doing benchmarks at speakeasy.net and then underestimating the values. On Mon, May 18, 2009 at 2:31 PM, Tom Eastep <teastep@shorewall.net> wrote:> Christ Schlacta wrote: >> I QOS filtered an entire subnet to use half my max upload, and low >> priority. my other subnet is configured to have up to 100% and high >> priority. >> >> wan0 1 4*full/10 full 1 >> default,tos-minimize-delay >> wan0 2 4*full/10 full 2 >> tos-maximize-throughput >> wan0 3 2*full/10 5*full/10 3 >> >> >> traffic ruled into mark 3 in tcrules includes a dedicated bittorrent >> server. for some reason, it still manages to slow category 1 and 2 to >> a crawl :( what might I be doing wrong ? > > These problems are often caused by using bandwidth settings that are not > realistic (e.g., what your ISP says your bandwidth is rather than what > it actually is). The settings in /etc/shorewall/tcdevices should always > be very conservative. > > Without details, we really can't advise you further. Please see > http://www.shorewall.net/support.htm#Guidelines. > > -Tom > -- > Tom Eastep \ When I die, I want to go like my Grandfather who > Shoreline, \ died peacefully in his sleep. Not screaming like > Washington, USA \ all of the passengers in his car > http://shorewall.net \________________________________________________ > > > ------------------------------------------------------------------------------ > Crystal Reports - New Free Runtime and 30 Day Trial > Check out the new simplified licensing option that enables > unlimited royalty-free distribution of the report engine > for externally facing server and web deployment. > http://p.sf.net/sfu/businessobjects > _______________________________________________ > Shorewall-users mailing list > Shorewall-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/shorewall-users > >-- (\_/) This is Bunny. Copy and paste Bunny (='.'=) into your signature to help him gain (")_(") world domination. ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects _______________________________________________ Shorewall-users mailing list Shorewall-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/shorewall-users
Christ Schlacta wrote:> http://bitch.aarcane.info/shorewall_dump > also, I''ve acquired the numbers in tcdevices from doing benchmarks at > speakeasy.net and then underestimating the values.From that particular sample, there was almost no traffic in classes 11 and 12: class htb 1:11 ... Sent 6588 bytes 122 pkt (dropped 0, overlimits 0 requeues 0) --- class htb 1:12 ... Sent 19999 bytes 42 pkt (dropped 0, overlimits 0 requeues 0) -- class htb 1:13 ... rate 600000bit ceil 1000Kbit ... Sent 15511984 bytes 22103 pkt (dropped 0, overlimits 0 requeues 0) ----- rate 757912bit 140pps backlog 0b 0p requeues 0 And the rate on class 13 is well below the limit. So it doesn''t look like there was any performance problem during the 3 minutes covered by that dump. -Tom -- Tom Eastep \ When I die, I want to go like my Grandfather who Shoreline, \ died peacefully in his sleep. Not screaming like Washington, USA \ all of the passengers in his car http://shorewall.net \________________________________________________ ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects
I've been tweaking on it today, and had kept the p2p hosts limited using their internal limiters until today, so I'll have to let it run and gather more stats while I can. On Mon, May 18, 2009 at 3:27 PM, Tom Eastep <teastep@shorewall.net> wrote:> Christ Schlacta wrote: >> http://bitch.aarcane.info/shorewall_dump >> also, I've acquired the numbers in tcdevices from doing benchmarks at >> speakeasy.net and then underestimating the values. > > From that particular sample, there was almost no traffic in classes 11 > and 12: > > class htb 1:11 ... > Sent 6588 bytes 122 pkt (dropped 0, overlimits 0 requeues 0) > --- > class htb 1:12 ... > Sent 19999 bytes 42 pkt (dropped 0, overlimits 0 requeues 0) > -- > class htb 1:13 ... rate 600000bit ceil 1000Kbit ... > Sent 15511984 bytes 22103 pkt (dropped 0, overlimits 0 requeues 0) > ----- > rate 757912bit 140pps backlog 0b 0p requeues 0 > > And the rate on class 13 is well below the limit. So it doesn't look > like there was any performance problem during the 3 minutes covered by > that dump. > > -Tom > -- > Tom Eastep \ When I die, I want to go like my Grandfather who > Shoreline, \ died peacefully in his sleep. Not screaming like > Washington, USA \ all of the passengers in his car > http://shorewall.net \________________________________________________ > > > ------------------------------------------------------------------------------ > Crystal Reports - New Free Runtime and 30 Day Trial > Check out the new simplified licensing option that enables > unlimited royalty-free distribution of the report engine > for externally facing server and web deployment. > http://p.sf.net/sfu/businessobjects > _______________________________________________ > Shorewall-users mailing list > Shorewall-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/shorewall-users > >-- (\_/) This is Bunny. Copy and paste Bunny (='.'=) into your signature to help him gain (")_(") world domination. ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects _______________________________________________ Shorewall-users mailing list Shorewall-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/shorewall-users
>also, I''ve acquired the numbers in tcdevices from doing benchmarks atspeakeasy.net and then underestimating the values. This is a timely post, based on my my recent attempts and ultimately successes in setting up QOS/TC. Any corrections to this information are appreciated. You may want to re-think your strategy of using Speakeasy ratings as a basis, even if you are conservative. My home cable connection, through a local rural ISP, shows 118kbps upload speed with Speakeasy. I found the real speed to be more like 12kbps (yes 12). On an organizational firewall, on a 3MB DSL connection, I found the "true" upload speed to be about 70kbps. Speakeasy here would show it to be a lot faster. Remember, you are trying to prevent queueing on the local interface (i.e. DSL or cable modem), so you need to determine the rate that achieves that. What I did was this: 1. During a time that the line was congested, I set the upload rate in tcdevices extremely low, where I was guaranteed that I wasn''t queuing. I would imagine on a decent internet connection you could start around 25kbps. It might even be better to put some artificial load on the line during a slow usage period so that you can be certain that the line congestion continues throughout your testing. 2. *Run shorewall restart && sleep 10 && shorewall show tc. Look at the output to see the effect of the new settings on your traffic flows, and also to verify that the line is still under heavy use. It should show traffic close to the value you set. 3. Determine if the connection is still congested (I used an RDP session from the internet to a machine on the network - I could instantly tell if the modem was queueing based on the latency of the response. I imagine you could do the same with loading web pages etc.) 4. Bump the upload rate in tcdevices slightly. 5. Repeat steps 2-4 until you find the rate that causes queuing & latency. I would run a few tests around that number to verify that the traffic is still at a level that would cause the connection to slow down without QOS. 6. Set your upload rate in tcdevices to slightly less than that rate. This process worked well on a 3MB DSL connection that hosts email and was subject to frequent bouts of incredibly bad real time internet responsiveness. Since getting QOS set up in Shorewall, they haven''t had a single incidents of "internet slowdowns." - Bob *Is there a better way to reset the TC counters than restarting Shorewall? ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects
Robert K Coffman Jr. -Info From Data Corp. wrote:> > *Is there a better way to reset the TC counters than restarting Shorewall?Not that I''m aware of. Thanks, Bob! -Tom -- Tom Eastep \ When I die, I want to go like my Grandfather who Shoreline, \ died peacefully in his sleep. Not screaming like Washington, USA \ all of the passengers in his car http://shorewall.net \________________________________________________ ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects