Tcp/ip Tuning For Better Speed
i have this couple of windows 2003 servers, colocated in data center, i need to improve download speeds to our customers who are at least 200ms away, the end user is not using download accelarator,
is there a way that any settings to be done on server so that per thread speed can be increased, this case the server and client both have the ability to make a connection at more than a megabit speed. i did some search but all the articles point to end user and not the server saying to increase tcp window size etc.. not sure if those articles relate to server side changes.
View Complete Thread with Replies
Related Forum Messages:
as part of a project I have lately been looking into various aspects of kernel tuning. Most notably lately tuning the TCP stack for more efficient memory usage/throughput.
Thought I would start this thread to mention some of the tools I'd found for doing testing and see what anyone else had to recommend.
So far my favorite of the bunch is nuttcp. Its easy to use and gives a very good idea of how much of your bandwidth you are able to utilize.
A few interesting web pages are as follows for anyone interested in the topic:
[url]- Tuning TCP for High Bandwidth Delay networks
[url]- TCP Tuning Cook book, some interesting information in there as well
[url]...formanceTuning - Performance Tuning TWiki. Has a list of useful tools, flags for existing tools and ways to monitor network performance from a system level, along with some suggestions of things to correct
I have Linux server with WHM/Cpanel with 2000 domains now my problem is.Mysql is using 90-100% CPU usage and 1500-2000 queries are running at a time so please guide me how can i optimize it and how can i tune mysql server so it doesn't go high.
I have configure my.cnf file as --->
max_allowed_packet = 4M
set-variable = max_connections=100
long_query_time = 3
how can i do this and any changes is require ?
Tuning Sites For MySQL
What have you found to be the best tuning sites for MySQL?
I'm getting into a bit of trouble. We have a weather site, and with all of the traffic, we're getting a little tapped out. When the loads hit between 134 and 160, the mail clients start to time out. Apache is still pretty fast, although it takes a little longer once you cross loads of 80, 5 second page loads, but when it hits between 130 and 160, I'm seeing 15-20 second page loads. DA is impossible above 80 but SSH is still very workable. Apache is tweaked to the max. I've kicked up some of the sizes in MySQL several weeks ago, and that did it them. However, we're taking on about 22,000 to 25,000 uniques an hour now. We normally can handle that no problem, but people are asking for maps a lot more now with the flooding and all. That requires a lot of MySQL lookups and the CPU creating a lot more maps. The maps I already cache for the duration, which is 15 minutes. The only horse I have left to whip is MySQL. After that, it will probably be a move to FreeBSD 7, but I'd like to throw in a few tweaks yet before we do that.
Tuning-primer.sh Results, How To Act On Them
MySQL Version 4.1.22-standard i686
Uptime = 0 days 0 hrs 4 min 15 sec
Avg. qps = 17
Total Questions = 4479
Threads Connected = 1
Warning: Server has not been running for at least 48hrs.
It may not be safe to use these recommendations
To find out more information on how each of these
runtime variables effects performance visit:
Current long_query_time = 10 sec.
You have 1 out of 4491 that take longer than 10 sec. to complete
The slow query log is NOT enabled.
Your long_query_time may be too high, I typically set this under 5 sec.
Current thread_cache_size = 128
Current threads_cached = 6
Current threads_per_sec = 0
Historic threads_per_sec = 0
Your thread_cache_size is fine
Current max_connections = 2000
Current threads_connected = 1
Historic max_used_connections = 7
The number of used connections is 0% of the configured maximum.
You are using less than 10% of your configured max_connections.
Lowering max_connections could help to avoid an over-allocation of memory
See "MEMORY USAGE" section to make sure you are not over-allocating
Max Memory Ever Allocated : 96 M
Configured Max Per-thread Buffers : 10 G
Configured Max Global Buffers : 58 M
Configured Max Memory Limit : 10 G
Total System Memory : 3.95 G
Max memory limit exceeds 85% of total system memory
Current MyISAM index space = 78 M
Current key_buffer_size = 16 M
Key cache miss rate is 1 : 735
Key buffer fill ratio = 8.00 %
Your key_buffer_size seems to be too high.
Perhaps you can use these resources elsewhere
Query cache is enabled
Current query_cache_size = 32 M
Current query_cache_used = 4 M
Current query_cach_limit = 1 M
Current Query cache fill ratio = 14.83 %
Your query_cache_size seems to be too high.
Perhaps you can use these resources elsewhere
MySQL won't cache query results that are larger than query_cache_limit in size
Current sort_buffer_size = 2 M
Current record/read_rnd_buffer_size = 256 K
Sort buffer seems to be fine
Current join_buffer_size = 1.00 M
You have had 0 queries where a join could not use an index properly
Your joins seem to be using indexes properly
OPEN FILES LIMIT
Current open_files_limit = 10000 files
The open_files_limit should typically be set to at least 2x-3x
that of table_cache if you have heavy MyISAM usage.
Your open_files_limit value seems to be fine
Current table_cache value = 1024 tables
You have a total of 721 tables
You have 93 open tables.
The table_cache value seems to be fine
Current max_heap_table_size = 16 M
Current tmp_table_size = 32 M
Of 212 temp tables, 0% were created on disk
Effective in-memory tmp_table_size is limited to max_heap_table_size.
Created disk tmp tables ratio seems fine
Current read_buffer_size = 1 M
Current table scan ratio = 17754 : 1
You have a high ratio of sequential access requests to SELECTs
You may benefit from raising read_buffer_size and/or improving your use of indexes.
Current Lock Wait ratio = 1 : 76
You may benefit from selective use of InnoDB.
If you have long running SELECT's against MyISAM tables and perform
frequent updates consider setting 'low_priority_updates=1'
how to make the changes in red? My server works good for awhile, but then gets REALLY REALLY slow.
Tuning A FreeBSD VPS For Network Performance
I have a VPS system on the west coast of the US, and access it from the east coast. Sometimes I can get 1Mbyte/sec downloads, and other times it is as bad a 250KB/sec.
I have done some pings, and have not seen any packet loss. I've experimented with sysctl and changed some parameters to hopefully help, but really haven't seen much of a difference.
Does anyone have a recommendation as to what I could do different to squeeze a little more speed out of the connection? The problem is that from both sides of the US, I see ping times (depending on different ISPs on the east coast) from 80ms-120ms.
Here is what I've tuned so far:
I'm trying to block all out going on ports that dont have a service binded to them to prevent back connects, and i have the following allowed in CSF:
ports with nothing binded to them:
not sure what these are but nothing seems to be listening on them so i removed them.
do i need 53 open on TCP?
whats 113 for?
cpanel? does it need to be allowed out?
UDP I have
20,21 in/out, are they needed?
Speed- At FDC ( Speed Tests )
I am having some serious speed issues with my 1Gbit server at FDC. After opening a ticket, they've simply dismissed it as a server configuration problem. However I am convinced it isn't because certain ISP's (usually universities) get good speeds, usually 700kb/sec but the vast majority of my users get between 20-50 kb/sec and it's causing a lot of complaints.
Furthermore I have other servers with FDC which are 100mbit which perform better than my 1 Gbit one. There are no server bottlenecks (CPU/RAM/HDD), since I've closely monitored them (PRTG) and they aren't even heavily utilised. So the problem is with the network at some point.
Speed Test : [url]
where abouts you are downloading from, your ISP and net connection. Wget's from servers are also welcome as are traceroutes.
KVM Over TCP-IP
I was wondering if this is a good solution for a server room or a colo facility where you have multiple servers:
what is a popular product that allows you to see messages on the server when it's
Too Much Packets To Tcp What Does It Indicate
What does it mean is it indicate ddos attacks?
From 18.104.22.168 - 1160 packets to tcp(1034,1036,1046,1055,1072,1084,1086,1097,1108,1109,1124,1138,1144,1146,1161,1174,1179,1180,1199,1206,1208,1237,1242,1275,1295,1296,1298,1313,1335,1 346,1349,1357,1384,1404,1419,1420,1475,1484,1509,1510,1538,1545,1547,1585,1593,1612,1684,1689,1690,1729,1731,1733,1736,1746,1749,1752,1753,1756,1762,1 763,1765,1768,1770,1779,1782,1784,1785,1786,1787,1789,1792,1794,1800,1806,1856,1877,1879,1885,1930,1988,2004,2005,2022,2027,2073,2077,2099,2109,2113,2 177,2178,2179,2180,2184,2185,2206,2237,2259,2266,2267,2282,2288,2313,2333,2500,2562,2565,2574,2585,2615,2617,2618,2657,2664,2666,2674,2686,2687,2808,2 821,2831,2836,2846,2867,2892,2904,2949,2950,2964,2984,2993,3101,3130,3210,3215,3285,3336,3359,3572,3638,3695,3696,3700,3848,3893,3973,4023,4030,4235,4 269,4293,4358,4370,4380,4398,4414,4472,4509,4549,4571,4585,4606,4608,4635,4685,4766,4778,4780,4812,4836,4844,4858,4902,4903,4909,4912,4916,4935,4936,4 937,4943,4955,4989,5534,5940,6245,6250,6256,6264,6367,7359,7564,7940,8538,9338,10203,10462,10763,11037,11332,11348,11462,11606,11633,11971,12177,12213 ,12242,12267,12276,12283,12307,12361,12399,12457,12472,12584,12645,12648,12793,12829,12842,12906,13197,13438,13807,14465,14493,14762,14765,14768,14769 ,14778,14779,14795,14981,15913,16474,16506,17060,17565,18047,18131,18191,18342,19113,20426,20702,21575,22062,22099,22379,22420,22423,22440,22675,22908 ,23100,23747,23766,24121,24248,24315,24365,24372,24411,24420,24425,24436,24486,24494,24639,25290,25507,26122,26702,26923,26975,27213,27302,27357,27409 ,27947,28731,28821,28982,29197,29227,29249,29285,29448,30472,30554,30564,30584,30632,31346,31628,31899,32074,32093,32306,32562,32566,32657,33968,33980 ,34442,34947,35047,35423,35599,35718,36937,38131,38404,38580,38696,38982,38995,38998,39001,39006,39036,39041,39077,39205,39288,39412,39822,39880,39999 ,40052,40942,41197,42090,42424,43419,43570,43991,43992,44917,46356,46515,46661,46669,46675,46814,46904,47594,48257,50086,50088,50316,50481,50511,50667 ,50786,50789,50790,50791,50792,50798,50802,50811,50930,50941,50951,50959,50999,51002,51008,51532,51650,51655,52362,52441,52448,52459,52531,52587,52612 ,53013,53223,53232,53237,53267,53284,53288,53941,54256,54789,55144,55228,55463,55522,55648,55846,56130,56807,57504,57765,57812,57814,58340,58850,59239 ,59945,60101,60150,60418,60648,60929,61313,61334,61431,61553,61733,61841,61848,61854,61857,61915,61921,61980,62035,62163,62403,62588,62899,62998,63081 ,63097,63198,63302,63379,63715,64214,64373,64380,64434,64442,64485,64491,64495,64501,64505,64514,65151)
Firewall TCP Out Connections
Firewall TCP Out Connections
My server started lagging up and I processed my configserver firewall logs and founds tons of TCP out connections. How can I track down which user was making these connections, if possible?
Tcp: Treason Uncloaked!
Lately my websites have been experiencing slow page loads at random times.
In my var/logs/messages I see this every so often:
Oct 17 11:31:06 app3 kernel: TCP: Treason uncloaked! Peer 22.214.171.124:63871/80 shrinks window 2128028156:2128028157. Repaired.
Oct 17 11:31:06 app3 kernel: TCP: Treason uncloaked! Peer 126.96.36.199:63872/80 shrinks window 2309432207:2309432208. Repaired.
Oct 17 11:31:07 app3 kernel: TCP: Treason uncloaked! Peer 188.8.131.52:63874/80 shrinks window 2635253661:2635253662. Repaired.
How do I stop these? Is there any other method besides banning IPs?
Ddos Tcp Window
Someone is ddosing us but i don't think is a huge ddos and i need your advise about that.
I own a dedicated server with direct admin.
2gb ram and 100mbit speed
DNS Tcp Connections
WARNING: One or more of your DNS servers does not accept TCP connections. Although rarely used, TCP connections are occasionally used instead of UDP connections. When firewalls block the TCP DNS connections, it can cause hard-to-diagnose problems. The problem servers are:
Error [No response to TCP packets].
APF is installed on the server, how do I allow TCP DNS connections? I already added port 53 to ingress/egress for TCP and UDP.
Corrupted Tcp/ip Stack
seems like my server has a corrupted tcp/ip stack. can it be resolved without actually reinstalling the OS?
also, what are the possible reasons that could have caused the corrupted tcp/ip stack, because it just happened suddenly out of nowhere =(
with a corrupted tcp/ip stack, there's no way i can remotely access it right?
the only way to fix the server is to either get the techs in the datacenter to do it or i'll have to go down personally to do it?
Grsec Kernel = No Tcp Traffic
Tried building a grsec patched kernel as I don't fancy getting that rootkit thats going around for a second time
it was linux-184.108.40.206 which was the latest one there was a grsec patch for.
compiled ok as a monolithic kernel however on reboot I was not able to get anything but a ping from the server, Figured that the kernel had paniced and not booted.
However the tech who rebooted the server for me and selected the other kernel (a standard centos kernel) said that the system had booted but was not responding to tcp traffic.
So its either a case of
1) I missed something important when configuring the kernel.
2) perhaps a problem with APF it does warn enabling Monolithic kernel support is unsupported in the config file (although it seems to work fine with the centos stock kernel)...
I'm wandering what the best next course of action is i'd quite like to be able to run a kernel that doesn't allow write access to /dev/kmem if possible.
Ip_masq_new(proto=TCP): No Free Ports
I have a rather serious issue that is causing us a lot of bother and embarrassment and would really appreciate some opinions on why it might be happening...
We have 5 linux servers in one of our racks protected by a Watchguard firewall. 2 web servers, 2 database servers and an smtp relay server.
This setup has been running without a hitch for well over a year now but both last night and the night before we have been getting the following message appear in the firewall logs:
ip_masq_new(proto=TCP): no free ports
This causes very strange results for our HTTP and HTTPS services, with images not loading and some ip based virtualhosts not working at all (although some still do). It seems other services like SNMP are not affected by this (even though they are NAT'ed too) although SSH stops working completely.
Both times this has happened a restart of the firewall has resolved the situation.
I've done a bit of research on the error and it seems to suggest that one of the internal servers is making an awful lot of outgoing requests although...
a) I cannot see any traffic spikes in our Cacti monitoring tool
b) How many requests is considered 'a lot'?
Pop :: Unable To Create Selectable TCP Socket
I have one mail server. and our one client sending and receiving mails through php coding so i have setup all the configurations in config.php file and sending mail ok but pop retrieves is not working and i m getting error------------>
Cannot create POP3 connection to pop.mxo2.com: Unable
to create selectable TCP socket (3494 >= 1024)
Measuring TCP Retransmits As A Symptom Of Packet Loss
a tool that can measure how much packet loss we are having on a given server by looking at the packets being sent from it. I.e, something than looks at all TCP/80 connections and measures how many packets and bytes are being retransmitted vs actual packets and bytes sent.
This documents explains it:
We need this to measure network performance of different hosts where we have dedicated servers. This would be a good way of measuring performance with the actual data of our users.
Does anyone know of such tool? I.e, something that can say
2532 packets/second - 132 retransmits/second (4.8%)
25.43Mbps/sec total traffic - 24.84 Mbps/sec actual data sent - 0.59Mbps retransmits
Even better if it can then break it out on IP prefixes. like
192.0/16: 0.2% retransmits
192.1/16: 3.2% retransmits
192.2.16: 0.3% retransmits
192.3/16: 22.5% retransmits
192.4/16: 0.3% retransmits
This would be a good indicator of connectivity between the host where the tests are ran and other specific hosts.
IWeb Technical Support Reps And System Admins Dont Understand Basic TCP-IP Networking
I want to pass on a warning before anyone goes with iWEB. I just ordered a $269 a month dedicated server so I could run vmware on it for my own virtual machines that I need.
With this server I got 7 IPs.
The primary IP they assigned me is say for example: 70.xx.15.171 with a GW of 70.xx.15.161. The secondary ips they issue me are 70.xx.18.249-254. So these are on a completly different subnets.
After talking on the phone to several differnt people including tech support who tells me this will work jsut fine useing thoes secondary ips on my VMs and to not put a gateway address. I talk to my account manager he says to open a ticket that he will forward to their system admins.
This is the ticket I opened.
I just got a new server set up CL-T113-140CL. It was
assigned the primary ip of 70.xx.15.171 with a GW of 70.xx.15.161. My
secondary ips are 70.xx.18.249-254. The server we got from you is a
VMWare host. We need to use the secondary IPs for the Guest VMs. How
ever the primary and secondary IPS are on different subnets and so
the VMs useing the IPs 70.xx.18.249-254 can not talk to the gateway.
I need to have a gateway server IP to use for my VMs.
Please either give me a valid GW address for the 70.38.19.x subnet or
assign new secondary IPs on teh 70.xx.15.x subnet with a SM that can
talk to the 70.xx.15.161 GW assigned.
this is the reply I get back from their system adminstrator
The gateway to be used for your secondary IP addresses is the same as
the one for your primary. You should not have anything to change,
gateway-wise. Just leave the current one as default gateway and
everything should work just fine.
If there is anything please do not hesitate to contact us.
Jean-Francois Doucet [iWeb]
System Administrator / Administrateur Système
Support / Client Hub : [url]
Are they brain dead at iWEB? DO they understand basic TCP-IP networking at all?
Here is my last reply. I tried to make it real simple for them to understand.
This does not address the issues. Do you even know what a virtual
machine is? It is a virtual computer running on the host computer.
The host computer has a valid ip address of 70.xx.15.171, sm:
255.255.255.224 gw: 70.xx.15.161. The virtual machines (or guest
operating systems) if I assign it the IP address of 70.xx.18.249 SM:
255.255.255.248 with no gateway it is physically impossible for it to
reach the gateway of the host pc 70.xx.15.161. You have to understand
the Virtual Machine operates as a completely separate computer. It
requires its own gateway. with no gateway specified it has no way to
get out of its local subnet to get to the 70.xx.15.x subnet. What
happens if you put two physical computers on one hub. The computer on
port 1 of the hub has the IP info 70.xx.15.171 SM: 255.255.255.224 GW
70.xx.15.161. The second computer is on port 2. It has the ip info
of: 70.xx.18.249 SM: 255.255.255.248 GW: NONE.
How is the pc on port 2 going to communicate with the pc on port
1? This is the exact situation we are in. There has to be a
gateway to route the IP traffic from pc2 to pc1 to get onto the
TCP Port "altport" Not Found
when i restarted exim from SSH i got below message
root@web[/etc]# /etc/init.d/exim restart
Shutting down clamd: [ OK ]
Shutting down exim: [ OK ]
Shutting down spamd: [ OK ]
Starting clamd: [ OK ]
Starting exim-altport: 2009-05-27 19:50:20 Exim configuration error:
TCP port "altport" not found
is there anything wrong with our serve
After reading tons and tons of messages on these forums... I have decided to get a VPS from KnownHost or SLhost.com.
KnownHost packages are a bit more expensive than SLHost (prepaying 6 months), however I am not looking to save a buck or two. I want the best speed for the money.
Can anyone tell me which of these two has the best performance (server performance and bandwidth performance) for the money?
200GB Premium Bandwidth
256MB Guaranteed RAM Burstable to 768 MB
10 GB Disk Space
$5 - Control Panel
$402 per year with Control Panel
300 GB Bandwidth
384 Guaranteed RAM (Burstable unknown?)
20 GB Storage
Free Control Panel
$389.64 per year
Just trying to figure out which one would give me the best performance...
How To Get Best Possible Speed.
ive got a site (heavily mysql/php based) and ive installed e-accel and mod_gzip. im wondering if anyone knows any other ways to speed up the site ?. mysql has been optimized also.
its pretty quick loading now but i want tho get it running as fast as possible
I am trying to purchase a server and HiVelocity gave me a really good quote. But their speed test did not turn out really good.
Anyone can share their views on their network and speed?
I want ask How i know the Server Download speed from Rapidshare ? I have windows Dedicated Server with 1GB Port and when i download something from rapidshare its just 2MB/s , is that Normal ? and now i am thinking to Buy another one , How i can know the Download speed from rapidshare site ? I ask some Company about that but No one give me test or any thing like that ,
I recently setup WAMP on my dedicated, and I'm unsure if the slow download speed is from WAMP or something else I need to remove from the dedicated. Might just be the distance I am from the dedicated because it's hitting 88.74Mb/s down and 71.19Mb/s up from a local city.
It's located in Kansas city, MO.
uploading some stuff to my servers, and it's going at about 50-60 kbps.
what would be a good service that could see my uploading speed rise to about 100-200 kbps?
I live in Canada, BC to be more specific
After reading a lot of good reviews on Softlayer around here, I was thinking about switching to them. They have some really good deals and I'm planning on getting one of their dual opterons, but I have a question about the processor speed in relation to database intense stuff (like a forum with a good number of users on at once, or a CMS).
I have the option of either going for a Dual Opteron 248 (2.2 gHz) or a 252 (2.6 gHz) which costs $50 more. My question: is the 252 really worth the extra money compared to the 248? All I have on my server is an IPB forum with 100-450+ users on at any given time, and a static site (soon to be converted to a CMS), and I'm thinking that RAM is more important than the processor for a database app like those (that, and hard drive speed). So I wanted to know if it would be better to go with the 248 and spend the money on more Ram and a SA SCSI 10k drive or two?
Speed Of My Site
I just bought a hosting on one of the American servers providers, so I am interested how fast my site loads at different locations around the world. Mainly I am interested in US, Europe and Australia. I am from Europe but, my connection is not so great
I want you input on how fast does it load on your end, and maybe how high ping is from your provider.
I hope there are some people who want to help me. I want post a link here because then people will complaint that I am advertising the site...
So if it is not a problem just post a message on this topic and I will send you a link true private message.
we have 9 rack each rack have 2 cisco switch 2950(1 for internet and 1 for private network) if we try to do a transfer files between two server we can't go over 1mbps, the same if we move the file from the web, every server nic is set to 100mbps the same for swith port, why we get a so slow speed?
We have 10 VPS node on one hyperVM.
One of the VPS have problem while uploading.
When I try to upload any contents to my site it gives me 10Kbps speed so I am worry about it.
I have one reseller account on which I get more than 200Kbps upload speed.
What can I do to fix this issue?.
if any changes I need to do on openvz.
I've tried googling, searching on webhosting talk for 100meg test file, speed test for linux server, ect.. ect.. ect.. and I cant seem to find a damn thing.
Anyone have a recommendation for a place to download a 100mb test file to test out my speed? I'm looking for a server on a gig line.
Speed Up NFS
how do we speed up NFS connections?
This is what I currently have in the server exports:
/images 220.127.116.11(rw,no_root_squash) 10.111.0.11(rw,no_root_squash)
Bandwidth Cost And Speed
My users frequently tell me that my website is slow, but it doesn't seem to be so, for me. Are there objective tools and criteria to test its speed (response time, max transfer etc)?
Also, I'm currently paying $1 for GB of transfer. How much does it usually costs?
Since my hosting company sets php_safe off, I'm considering changing it.Can you guys recommend me a hosting plan that has:
- ssh with vi etc: this is important
- A FAST server
- Norway-based (to enjoy .torrents without being bothered) or US Based (to enjoy "fair use", which seems to be exclusive to the USA)
- 1-3GB of space
- LOTS of transfer. I don't consume many gigabytes yet, but someday I will.
- Some kind of hacking protection. I'm damn scared of my website suffering a vampire attack and having to pay for the raeped bandwidth.
Limit Download Speed
im using vmware in centOS dedicated server
how i can limit Download speed for each IP ?
how i can limit download speed for each Vms?
how i can limit download speed for all Vms?
Corenetworks Speed Test
Anyone got a 100mbit server at corenetworks that I could speed test from? The test file that I got from corenetworks was only able to give me about 500KB/sec so I think their server might only be 10mbit.
How Can I Test Vps Speed
I ordered my first vps and uploaded a 1 page web site with a few graphics. I browsed to the web site and it took longer than normal to load the entire site with all graphics.
Is there a way that I can test my vps account speed?