Does increasing upload or download speed reduce ping?

Question

I have a 5 MB/s connection, but I always notice high ping when playing online games. When the network is stable (i.e. when pingtest.com reports the connection as 5 MB/s), the ping stays at 50 ms, but my previous connection was 2 MB/s, and it also had a 50 ms ping when stable.

Why isn't my current connection's ping 50 ms when it's running at 2 MB/s like my previous connection? And which makes more of a difference in ping, upload or download speed?

Answer 1

Latency is generally independent of bandwidth.

• Latency, as commonly measured by ping times, is an indication of how long it takes for your system to send data to another computer and to receive its response. Latency is influenced primarily by how far the data must travel—it takes much longer to access a webpage on a server that is halfway around the world than one that is on a server next door. Longer communication routes, where the data must pass through many switches and routers, increase latency as well.

• Bandwidth, usually expressed in megabits per second (Mbps), is the amount of data that can be sent or received by your system over the network per unit of time. Bandwidth is what you typically pay for when you purchase Internet service—with most home Internet services, more expensive plans provide more bandwidth.

• Although higher bandwidth can improve latency during times of congestion, under normal conditions, the two are generally independent. The HughesNet Gen4 satellite Internet service, for example, provides good bandwidth (up to 15 Mbps), but suffers from poor latency, averaging about 700 ms ping. This high ping is due to the nature of satellite Internet service, which requires beaming data up to a satellite and back down. Conversely, a dedicated T1 line has only 1.5 Mbps of bandwidth but can provide latency as low as 10 ms.

See also on Server Fault: Will a higher bandwidth Internet connection lower ping response time?

Answer 2

Consider how latency and bandwidth works with hard-copies carried by foot.

Say you have a boxful of hard copies—about as much as you can carry without being significantly slowed down by the weight, and two places you might want to bring it, one 50 m away and one .5 km away.

Let's say you walk at about 5km/h. It would take you 1.2 minutes to bring the box to the first place and return, and 12 minutes to bring the box to the second place and return.

Now lets say you have to bring 100 boxes. You'd have to make 100 trips, so your total time carrying would be 2hours and 200hours respectively.

Now, let's make these things better.

Let's say you enlist 99 helpers. You can now have more bandwidth. You can now bring the 100 boxes in 1.2 and 12 minutes respectively.

Let's say you enlist 2000 helpers. You now have even more bandwidth. You can now bring the 10 boxes in 1.2 and 12 minutes respectively. The extra bandwidth hasn't helped because you've maxed it out, and the latency is the same.

Let's say you get rid of the 99 helpers, but buy a bicycle, and you can do a healthy 40km/h on it. You can now do the one-box trips in 9 seconds and 1.5 minutes respectively. You now have less latency. The 100-box trips will take 150 minutes and 25hours though.

Now, obviously the bicycle (lower latency, some extra bandwidth) is better at bringing one box of papers quickly, while the massive team of helpers (same latency, much more bandwidht) is better at bringing lots of boxes.

Network connections compare with each other in ways similar to how these different means of transporting hard copies compare.

Downloading a very large file is akin to the task of transporting lots of boxes, and so the more bandwidth, the better.

Playing a game tends to involve lots of small messages, so it's like the task of carrying one box over and over (we can't bulk-transport all the boxes, as the next box isn't ready yet). The lower the latency, the better.

But in our analogy, there's no reason why we can't have a massive team of helpers who all have bicycles.

And extend the analogy, different network connections will differ not only in latency (foot vs bicycle) and bandwidth (how many helpers) but also in having different short-cuts available, and different points they have to go through, so one may be lower latency for one trip and higher for another.

But where the analogy is accurate is that while we can have two connections where one is better by one metric and worse by another, we can also have two connections where one is better than the other by both measures.

Answer 3

The correct answer is neither.

Ping can be done accomplished a couple ways, through UDP or ICMP (A short intro can be found here for ICMP)

The short of it though is that one machine sends a series of special packets, that are forwarded from router/ switches and through gateways if required to leave one network for another. This trip through devices is a ping time, which is encapsulated and sent back. Generally two pings do not have to follow the same route. The number of hops is the biggest determiner on average of ping times, although there are times when higher upload or download could matter (flooded networks).

Answer 4

Both are equally important, because ping measures the roundtrip time, as in, you send a packet to some machine (upload), and a packet is returned (download), and the time from start to finish is the ping time. As a result, both should play an equal part in your pingtime.

However, there are some things worth noting: Most home connections nowadays have lower upload rates than download, so upload can clog easier, resulting in a longer packet queue, hence higher ping (and/or packet loss).

Downloading only 2mb out of 5mb does still impact ping because, amongst other things: - There is a lot of overhead with downloading, especially with protocols such as bittorrent. In general, one should expect AT LEAST 20% overhead in addition to net download rate, due to necessities such as TCP Windowing, TCP Header, Ethernet Header and Footer, IP Header, etc. On bittorrent, this overhead is magnified greatly because of the large amount of connections in use, and the need to continuously establish new ones. - Buffers along your network route clogged up - Increased chance of dropped packets, requiring a resend.

It should be noted that there are many things referred to as ping. You have the most common one, ICMP Ping, you also have various implementation of a tcp equivalent, and most games roll their own to produce a form of UDP Ping. The latter can vary greatly, as everyone seems to have their own idea of how it should be implemented to reflect how the link quality should be measured for the game, such as Round Trip, or one way only?

On a related note, nowadays when home and server connections are normally "fast enough", the biggest impact on a healthy route between you and the server is the amount of network hops. This is normally a result of the physical distance between you and the server. The amount of network hops can be checked by doing a traceroute.

Answer 5

A big factor is how much unused bandwidth you have for uploads, and how much you have for downloads. Since ping works by sending a message and listening for a response, both directions affect the total round trip time. The packets are essentially the same size in both directions (unless something fragments or reassembles a fragmented packet). If either your download or upload pipe is almost full, ping is more likely to be delayed. Most home connections are asymmetric, which just means you have more download bandwidth available. As a result, it may be easier to fill up your upload bandwidth, depending on your usage. Often times, ICMP (ping) traffic has a lower priority than other traffic, so it gets delayed or dropped when the pipe gets busy.

Answer 6

Bandwidth (speed) is necessary to be capable of sending out pings and receiving responses.

However, pings are usually designed to be compact (typical ping packets are much less than 1 KB), so unless you are flood-pinging and/or use huge packets, or have some very exotic connection like an air-gap circumvention via sound, it does not matter in practice.

Latency (responsiveness) is caused by signalling and processing speed of all hardware and software components on the round-trip path and is almost always the limiting factor if you consider the relations.

For example, assume you can send or receive 1 MBit/s, or roughly 0.1 MB/s, and your ping is a whopping 1 KB, i.e. 0.001 MB. This means you could send or receive 1,000 pings per second, each one requiring 1 ms to send or receive if you fully use the connection.

Typical latencies are > 10 ms for a round-trip over the Internet.

In that time you could send 10 pings already before receiving any response over that rather low-speed link.