Does latency ever get lower?

[Aeonix]

Each years our computers are getting faster, games more demanding, which fuels better computers.
We develop more memory in lesser spaces.
We develop processors that calculate more than super-computers 20 years back.
We had 56k modems, 1M, 2M, 5M networking, just recently there is possibility to get 1.5G at your doorstep, with record of 70G+

Everything moves on-ward, but what about latency? The speed at which TCP/UDP package can be delivered?

Is that improved since "back then"? I can't see anything on the internet, so I'm wondering, since you guys have been here longer than I am.
If not or so, are there any hopes or idea of improvement?

"We know" how to improve processors.
"We know" how to improve memory.
"We know" how to increase networking bandwidth.
Do we know how to decrease latency? Do we care about enough it?

[Andrew_36]

This is a very open ended question as there are many factors that could lead to increasing the speed at which TCP/UDP packets get delivered. For example, in my opinion, increasing processor capacity and memory capacity in a service provider router would increase the time it takes for that device to process the TCP/UDP packets, so obviously the faster the devices get the faster processing of packets you get. The processing power the faster the router can compare TCP/UDP packets to policies, routing tables, ACL's, and QoS which would also affect latency. So my opinion is the faster CPU and memory that is implemented in these devices the better. However, with that said it is a very costly and timely process to upgrade a network. I work for a service provider and we are still using equipment from the 90's to early 2000's, and this is dealing with financial data. So even though the processing power is out there, it doesn't necessarily mean that it is being used. But there are other factors such as the physical medium used between devices and how far apart those devices are, the type of NIC you are using on a computer, etc.

[Aeonix]

While you provided good insight into situation, it's not quite what I hoped to receive.

Let's say that package from Netherlands to West U.S. takes 160ms, is there a chance in the future that the latency will go down? We're assuming latest of hardware, and no limits when it comes to money.

I'm talking about networking technologies development in very near future (talking about 5 years - 20 years), that could bring this thing down. Coax got upgraded to fibre to bring up higher bandwidth, that is a very good way to increase bandwidth and maybe reliability. It got implemented and people can enjoy it if they need it. Are there any "very good ways" that are expected in future that would bring down the latency? People knew about fibre before it got implemented, it was in plans, it took time to do, but now it's here. Is there something that we know about "latest latency breakthrough", if we even have any?

Rather than stamping out milliseconds by perfect hardware, I'm more wondering about technologies that could easily dump 20% of latency because of "The Greatest Revolution Ever" etc. etc.

[AssertNull]

I can't see anything on the internet

You didn't actually write this, did you? I won't post a Let Me Google That For You link... This time... :-)

We're assuming latest of hardware, and no limits when it comes to money.

No need to wait for new hardware. Stick your computer from 20 years ago on both ends. That ain't the holdup. Since money's no object, we can dispense with most of the other things that complicate things like the fact that we're not the only ones doing this. On our system, we're laying a nice long cable from the Netherlands to the US or perhaps we're using radio waves. The point is we're rich enough to have this cable or whatever all to ourselves and any other computers or other equipment along the way exists solely to help us get our message from point A to point B, no other consumers. We have a closed loop system, so no routing delays and we can dispense with firewalls, switching delays when it's not our turn, waiting for resources, and the entire TCP/IP protocol. We'll hire some super smart engineers to figure out the voltages, frequencies, etc. to transfer information long distances, which will hopefully be as close to the speed of light as possible, so do some math and figure out how long that is for the distance.

What will REALLY get way faster in your scenario is the transfer of a very large message as opposed to packets since we're the only ones doing this in a closed system.

In summary, no need to wait for new technology. It's here now. Latency significantly above the speed of light is due to factors other than our current technology.

[Reverend Jim]

It's here now

Well, almost. In the US, at least, the major providers have little incentive to actually make major improvements to infrastructure that would provide substantial improvements. They have taken billions of dollars that were intended to make these improvements and instead just kept them. In some cases they have actually blocked improvements in services by preventing some cities from providing services that would compete with the big (privately owned) ISPs. Why improve services when you can just keep a monopoly and charge what you want?

[Aeonix]

In such case, what is the potential change (or vague expectations) when it comes to latency. If they used those mentioned billions and actually did improve it?

[AssertNull]

I think my main point is that you used "we" a lot in your posts, as if we're all in this together reaching for the same goal of faster load times and fewer delays for everyone and cooperating for a single shared goal like the moon landing when in fact, we're not. Lots of folks have competing interests in all this and don't give a damn about YOUR latency, only theirs. Sounds like that's Jim's point too. The entire US has a ridiculously obsolete power grid. No new technology is needed to fix it and new technology WON'T fix it. Ditto latency. Websites don't load slowly due to lack of technology. The corollary to Moore's Law is that every time technology improves, people will make everything more complicated so we'll never get the fill benefit of the improvement.

[rubberman]

There is always the speed of light to consider, plus switching time in the necessary optical cable repeaters. That's about 135ms to traverse the world without repeaters, routers, switches, etc. From Amsterdam to NYC you still are looking a 20ms just to cover the distance. There there are all of the routers, spy systems, switches, etc. Longer if you go via satellite. These are barriers that cannot be avoided. So, if your packet from Amsterdam to NYC takes less than 100 ms, consider it a win. Try pinging a server across the Atlantic like that with Speedtest.net to see what is your likely best bet.