Tag Info

Hot answers tagged

87

Speed of Light: You are not going beat the speed of light as an interesting academic point. This link works out Stanford to Boston at ~40ms best possible time. When this person did the calculation he decided the internet operates at about "within a factor of two of the speed of light", so there is about ~85ms transfer time. TCP Window Size: If you are ...


36

For me, setting the ServerName property in httpd.conf fixed the delays (they were up to 10 seconds at worst): # ServerName gives the name and port that the server uses to identify itself. # This can often be determined automatically, but we recommend you specify # it explicitly to prevent problems during startup. # # If your host doesn't have a registered ...


31

This site would suggest around 70-80ms latency between East/West coast US is typical (San Francisco to New York for example). Trans-Atlantic Path NY 78 London Wash 87 Frankfurt Trans-Pacific Path SF 147 Hong Kong Trans-USA Path SF 72 NY Here are my timings (I'm in London, England, so my West coast times are higher ...


21

What does "latency" mean ? Are you comparing ICMP echo response times with application processing round-trip times ? If you are, that would explain it - and also tell you not to compare apples and oranges :)


19

I'm sorry to get here so late, but there is a book from Ilya Grigorik called "High Performance Browser Networking" that answers exactly this. There is a whole chapter (7th) dedicated to mobile networks. The book states that the problem with high performance is almost always tied to latency, we usually have plenty of bandwidth but the protocols gets in the ...


17

The Short Answer Yes The Long Answer It may not necessarily need to be right in the country you are targeting - just a close region could be sufficient. A good start is to determine where the major internet backbones/intercontinental links are in relation to your target audience, and plan accordingly. This way you could position services in a more ...


17

The issue was with Apache's main settings file httpd.conf. I found this: There are three ways to set up PHP to work with Apache 2.x on Windows. You can run PHP as a handler, as a CGI, or under FastCGI. [Source] And so I went into the Apache's settings and saw where the problem was: I had it set up as CGI, instead of loading it as a module. This caused ...


11

I had the very same problem. Setting localhost redirect to 127.0.0.1 in hosts file did not help. Optimizing MySQL server did not help (InnoDB -> MyISAM, changing many cache related directives in my.ini). Then I used web webgrind and narrowed down the problem to "new PDO(...)" call. Changing mysql:host=localhost;dbname=dp-ui;charset=utf8 to ...


11

Generically, you can use some of the advanced switches to the iperf utility to get a view of the network performance between systems, specifically latency and jitter... Is this a UDP or TCP-based message stream? I commented above on needing more information about your setup. If this is a low latency messaging application, there's a whole world of tuning ...


10

Traceroute involves sending UDP packets to each node along the way, and waiting for its timeout response (then moving on to the next node), whereas a ping is just forwarded. What you're seeing is the time it takes for each node to respond to the request instead of just forwarding a small packet. This is a pretty nice explanation of the whole process, and ...


10

From Key West to the topmost Northwest Passages Islands is about 4,000 miles which is a theoretical minimum of 20 milliseconds. Your theoretical minimum roundtrip time is 40 ms. I think you can safely remove the boundaries of physics from the equation. If you're in Toronto, then most of your customers will be much closer to you. What you need to accomplish ...


10

Measure with ICMP first if at all possible. ICMP tests typically use a very small payload by default, do not use a three-way handshake, and do not have to interact with another application up the stack like HTTP does. Whatever the case, it is of the utmost importance that HTTP results do not get mixed up with ICMP results. They are apples and oranges. ...


9

It will usually take more then those two options. Ping measures just the time from client, to server, and back again (rtt - round trip time) Usually databases use TCP, so you first need to send a SYN packet to start the TCP handshake (to simplify let's say 15ms* + cpu time, then you recieve and SYN/ACK (15ms+cpu time), send back an ACK and a request ...


9

Clearly you have an interference problem. Interference can come from passive elements like aluminum wall studs or thick floors, but those are not likely to show the periodic pattern you see. So something electric or electronic is periodically emitting. Finding it may be expensive or tough, but you have a few options. Graph more. It would be nice to make ...


8

can it be you have packet losses [that cause reteransmissions] for instance for larger packets? maybe try ping -s 1400 address under linux or ping -l 1400 address under windows. if that does not help try looking at the traffic with wireshark - maybe there are some strange re-transmissions, corrupted packets?


7

You won't get this information just on one click. What you can try is to use many different Looking Glass servers and based on the information you get there you get some estimates. Here is a list of some looking glass servers. Another possibility is to use some traceroute servers. I recommend you to google for e.g. some famous newspapers in the regions you ...


7

As you probably already know, DAVG refers to disk latency, and yeah, greater than 30msec is usually going to give you a noticeable decrease in performance and responsiveness. Latency can be caused by a lot of issues but first and foremost your disks must be able to handle the IO load you are throwing at them. IO load refers not only to the # of IO's per ...


7

Taskset is for binding a process to one or more CPUs; essentially specifying where it can run at initial execution or while it's running. If using RHEL/CentOS on modern server equipment, numactl is recommended over taskset. Cpuset/cset is for CPU shielding and is a framework build around Linux cgroups. Cset was never popular on certain distributions (like ...


7

A router's job is to route packets. It's not a ping responder. It can route packets just fine even if it can't respond to pings. Since you're seeing latency of less than a hundredth of a second to the hop after it and no packet loss, I'd say it's routing just fine. How can hop 7 be smaller than 6 and 8 smaller than 7 and 6!?? Shouldn't the pings be ...


7

Nope, a VPN can't alter physics to change the speed of light.


6

First, "speed" means nothing. You need to specify latency or throughput as requirements (one, the other, or both). This question is going to be almost impossible to answer. It would depend very much on the connection you have, who it peers with, etc. The raw distance from the US to Canada isn't going to make much difference in the latency (unless you're ...


6

What could cause this is the design of the SMB protocol. SMB requires a lot of acknowledgemenst, in some case between every single operation. This requirement for lots of acknowledgements means that your latency will kill your performance. It is block based instead of stream based. See: http://blogs.technet.com/b/neilcar/archive/2004/10/26/247903.aspx ...


6

In regards to kernel tunables for latency, one sticks out in mind: echo 1 > /proc/sys/net/ipv4/tcp_low_latency From the documentation: If set, the TCP stack makes decisions that prefer lower latency as opposed to higher throughput. By default, this option is not set meaning that higher throughput is preferred. An ...


6

In my experience the biggest cause of abnormal latency on otherwise healthy high-speed networks are TCP Windowing (RFC1323, section 2) faults, with a closely related second in faults surrounding TCP Delayed Acks (RFC1122 section 4.2.3.2). Both of these methods are enhancements to TCP for better handling of high speed networks. When they break, speeds drop to ...


6

If the box has enough CPU/memory/bandwidth, nothing at all; almost every modern computer couldn't care less about handling routing for a 100 MBit network, unless you really throw in lots and lots of rules.


6

All other things equal, you will have additional 44 milliseconds of latency just because of the speed of light. Give or take 1/20 of a second for each packet roundtrip. Not much for typical web usage. Passable for ssh sessions. Substantial if you access your DB directly with a lot of small consecutive transactions. I've ignored extra latency caused by ...


6

I'm seeing consistent differences, and I'm sitting in Norway: serverfault careers 509ms 282ms 511ms 304ms 488ms 295ms 480ms 274ms 498ms 278ms This was measured with the scientific accurate and proven method of using the resources view of Google Chrome and just repeatedly refreshing ...


6

This sounds like you're still in the design stage. If that's true, a few things to look at are: NFSv3/4 over NFSv2, to allow for larger packets, and features like "safe asynchronous write" Check your NFS client for read-ahead and delayed write, both of those features will help Obviously keep network latency low - GBit connections over a fast switch Make ...


6

There is a distance delay and all other things being equal (routing efficiency, processing overhead, congestion, etc.) a site on the west coast accessed by a host on the east coast is going to take longer than if that site is on the east coast but we're talking milliseconds here. Here's a nice article that explains latency and jitter: ...


5

If you wanted to just have a look on the bandwidth actually used, give nload a shot. I always prefer testing the daemon which is serving clients (wget/curl when testing a webserver, lftp for ftp-servers, etc). Artificial tests like iperf are better to check the general throughput of your routers, switches, NICs and IP stacks. HTH, PEra



Only top voted, non community-wiki answers of a minimum length are eligible