Questions about Ping/Reducing Ping

[Morgray]

I have a few questions regarding ping, which I hope to get answered on these forums.

For an environment in which every nanosecond counts, I´d love to find ways to reduce the time between transmission and reception of data (by using ping as an example/proxy)

First of all, I am not quite sure about what a basic ping actually includes. Is it just the time it takes for a package to be sent from my computer and received at a different computer, or does it also include the time it takes for the receiver to sent the message back to me, the originator of the initial ping?

Also I`d like to know wheter or not in every ping the access time of a regular HDD is included, seeing how HDD access times usually range anywhere from 2-12~ ms and if so, would using a RAMDISK/SSD get rid of the access time overhead and thus decrease ping time?

My last question would be how much upload- and download speed effects ping time.
Seeing how a basic command line "ping" command uses 32 bytes, I can only assume that the faster the upload/download speed is, the more beneficial it would be for the amount of time saved.

If anyone has answers and maybe even further suggestions to reduce ping time I would greatly appreciate it.

Thank you.

 

[JackMDS]

You "Bundling" together under the word Ping different occurrences.

Pure ping has Nothing to do with HDD they are just using RAM and Network .

If you are including under ping Internet connection than local ping are usually Negligible.

In other word try to explains what exactly you are doing so we can know whats is the actual variable to deal with.

P.S. To separate local ping from Internet timings run a Ping test between your computer and compare it to the one that you can measure with an App to a web site that represent what you are dealing with.

Free and very good portable Ping App - http://www.nirsoft.net/utils/multiple_ping_tool.html

 

[Morgray]

You "Bundling" together under the word Ping different occurrences.

Pure ping has Nothing to do with HDD they are just using RAM and Network .

If you are including under ping Internet connection than local ping are usually Negligible.

In other word try to explains what exactly you are doing so we can know whats is the actual variable to deal with.

P.S. To separate local ping from Internet timings run a Ping test between your computer and compare it to the one that you can measure with an App to a web site that represent what you are dealing with.

Free and very good portable Ping App - http://www.nirsoft.net/utils/multiple_ping_tool.html

Thank you for you quick answer.

What will be done is using a Trading Program to make transactions over a distance of around 1750 miles over the internet. Time is essential and the less time it takes the better. So I am looking to find ways to reduce any lag/make it faster on my end.

I thought using a ping as an example would fit since the data amount for the transactions is fairly small probably less than 250kb.

I hope that helps

 

[dave_the_nerd]

I have a few questions regarding ping, which I hope to get answered on these forums.

For an environment in which every nanosecond counts, I´d love to find ways to reduce the time between transmission and reception of data (by using ping as an example/proxy)

First of all, I am not quite sure about what a basic ping actually includes. Is it just the time it takes for a package to be sent from my computer and received at a different computer, or does it also include the time it takes for the receiver to sent the message back to me, the originator of the initial ping?

NIC 1 to NIC 2 to NIC 1. Round trip.

Also I`d like to know wheter or not in every ping the access time of a regular HDD is included, seeing how HDD access times usually range anywhere from 2-12~ ms and if so, would using a RAMDISK/SSD get rid of the access time overhead and thus decrease ping time?

Nope. Ping time is strictly NIC to NIC.

My last question would be how much upload- and download speed effects ping time.
Seeing how a basic command line "ping" command uses 32 bytes, I can only assume that the faster the upload/download speed is, the more beneficial it would be for the amount of time saved.

Nope. Ping is a measure of latency - the packets are super-small so that bandwidth doesn't effect them much, or at all, really. You can definitely have high bandwidth and high latency (satellite internet connections are famous for this) or low bandwidth and low latency (hard-wired serial ports.)

Different service tiers of the same technology (1.5, 5, 10, 20, and 40Mbps cable modem service) will all have similar latency.

 

[imagoon]

Thank you for you quick answer.

What will be done is using a Trading Program to make transactions over a distance of around 1750 miles over the internet. Time is essential and the less time it takes the better. So I am looking to find ways to reduce any lag/make it faster on my end.

I thought using a ping as an example would fit since the data amount for the transactions is fairly small probably less than 250kb.

I hope that helps

Dedicated point to point would lower the ping times the most.

Barring that, anytime you get the Internet involved it becomes a crap shoot.
Low latency connection directly to a peering provider. (Think XO to Level3)
Connection to the same peering provider serving the other end. (Think Level 3 if the trading system is attached to Level 3.)

If you are using some thing like Comcast or Time Warner "Business Class" you basically have no control what happens. If you plug in to their Metro Ethernet connections you have better odds to get lower latency.

Basically to have any real control over latency you will need to pony up for a good Internet connection. A good local router also helps. Anything "home office" doesn't count.

 

[QuietDad]

Your ping and latency can change at a moments notice too. If you do a tracert over the internet, you'll see there are many intermediate hops between two sites. On any given day the path can change, sometimes better, sometimes worse. If it's THAT time critical, only a direct point to point line will do what you need.

 

[azazel1024]

Is this just a class/theory, because if that is what you are trying to do, involve an actual consulting company. Trading applications where every nanosecond counts are BILLION dollar enterprises, not something you should be designing based on some forum suggestions.

 

[matricks]

If you don't control the weakest link between the two points there isn't a whole lot you can do, but there are some things you can do to at least ensure the links you control work as well as they can, considering your goal.

Bufferbloat.net may have some interesting information for you. Most (if not all, I'm not deeply involved) of the improvements the CeroWrt project have made have been pushed upstream to OpenWrt (in particular the sqm-scripts package), so your choice of hardware is wider than the scope of CeroWrt. It doesn't cost a whole lot to try it out, but with this use case I would think there is some budget room for more enterprise-like hardware suited to your use case (and professionals to go along with it, as suggested by azazel1024).

 

[JackMDS]

The pings on your computer are almost meaningless as compare to the one generated on the Internet routes.

If you have a specific Trading provider that you are using consult with them they might have specific solutions for their service's Routes.

General example - https://www.lightspeed.com/knowledge-center/top-faqs/faqs-technical/

Here is an article that shows how serious is this issue and beyond end user control.

http://www.extremetech.com/extreme/...-cost-of-cutting-london-toyko-latency-by-60ms

 

[Fardringle]

What are you doing where "every nanosecond counts"?

Most of the time you won't have any control at all over your PING times, particularly if you are communicating with servers across the Internet.

A PING involves your computer or device sending a request to a remote computer or device. The PING time is the total time elapsed between the time when your computer sends the request and when it receives a reply.

Yes, the remote device itself can have some very tiny effect on the PING time. If the device is overloaded then it might take slightly longer to receive the reply. However, the vast majority of the result is the time that it takes for the request to transmit from your computer to your default gateway, to your ISP, to their tier/host provider, across the host's backbone connections to the next steps on the path, and eventually to the destination, and then back again.

Your upload and download speed will have virtually zero effect on PING times unless you are overloading your connection with very large uploads and downloads (Torrents and other P2P file sharing, for example).

 

[Enigma102083]

If every nanosecond counts, then you should NOT be sending your packets 1750 miles. You should have your trading application co-located 1750 miles from you.

 

[Morgray]

Thanks to all of you for your wonderful suggestions. You have definately pointed me towards helpful directions I need to learn more about. Knowing where to look into/learn about is critical .

Yeah my "every nanosecond counts" statement was a "bit" exaggerated, it even made myself cringe that I phrased it like that. Every millisecond is more like it