Bandwidth Speed, Up / Down and all around

Are you confused about bandwidth, throughput, pipe, upload, download, and the “big words” the “engineers” like to throw out there like; asynchronous, synchronous or “DIA” (Dedicated Internet Access) and how it relates to Internet connection speeds?  You might be surprised that even the “engineers” don’t understand it either and are just as confused as you are (believe it or not).

Remember that bandwidth is bandwidth is bandwidth…is bandwidth.  If you have 10 Mb/s (that’s “Megabits per second” which is different than “MegaBytes per second” – Big “B” vs small “b”) and you’re downloading or uploading 9 Mb/s you still have 1 Mb/s to use in either direction (in theory).  You’d think this math is easy to understand right?  At such low speeds (10 Mb/s) there’s a little gray area and we don’t dive into those weeds,  however; at higher speeds there is less gray area on what you’d expect to see as a tolerance for acceptable loss.  At 30 Mb/s you’d expect to be at least near the 27 Mb/s speed on a download and upload on a synchronous connection (30 up / 30 down).

I recently had a run-in with an “engineer” at a large regional ISP in my neck of the woods who blathered on about asynchronous vs synchronous connections and how that if you’re experiencing problems with bandwidth it’s because off-site replication is “getting in the way” of other uploads or downloads like the kind you see during daily traffic.  That’s not a perfect quote but pretty darn accurate.  This leads me to note that…I hate the fact that it seems everyone these days is a “network engineer” or calls themselves an “engineer” when a lot of these people can’t seem to engineer their way out of a paper bag.  It’s a wonder how these people even got jobs doing what they’re doing!  All engineers are not created equally, especially when this engineer who’s one of the most arrogant I’ve run into in a long time knows very little about such a basic topic.  From one engineer to another you can’t dumb things down to uber simplicity like “getting in the way”…that just doesn’t hold technical water and it makes YOU look like the derp.

Why?  Why does this engineer know so little about such a basic topic?  I don’t happen to know but the derpgineer (that’s a derpy engineer) should understand that bandwidth is bandwidth is bandwidth as I started out.  It just is!  I don’t know how else to explain it, it’s such a simple concept.  You either have 30 Mb/s or you don’t!

This derpgineer also yammered on telling me how a tool called iperf works.  He went on to try and tell me how “ipperf” works saying that it just blasts UDP packets onto the network, blah blah blah…I heard blah blah blah after he mentioned “UDP packets” because anyone who’s used iperf for testing knows that the tool is not called IPperf, it’s “iperf” and that by default it uses TCP not UDP.  He trainwrecked midway through his first sentence!  I didn’t get the chance to enlighten him on the conference call because I was interrupted by the moderators.  Ugh…I don’t struggle with someone who knows less than me or even someone who knows more than me but when you fake know something…that’s when I get irritated!  It’s not called “IPperf” nor does it use UDP by default…the tell tale sign of a derpgineer is speaking about something authoritatively when they should go RTFM.  The ultimate issue is the all knowing all seeing derpgineer wasn’t even listening to what I was saying or the data that iperf put out which led to my questions for him.  It’s not…like…my opinion man!  The derpgineer just wanted to be right with less visibility into the bandwidth issue than I had.  I had more visibility than they did and I wasn’t being heard.  Most importantly, the data wasn’t listened to…it’s not about me or personalities it’s about the data!  It’s so frustrating to work with someone who doesn’t listen and pretends to know what they’re talking about!

Now…

On a 30 Mb/s connection they were arguing that replication was taking 20 Mb/s and since I was getting a 355 Kb/s transfer speed with something unrelated that we need to crank up our Internet speed (bandwidth) to something greater than 30 Mb/s.  Nothing could be further from the truth in fact!  If you have 30 Mb/s upload (and lets assume download too – 30 down / 30 up <- synchronous) and so do they (the derpgeineers ISP)…then I have (30 Mb/s – 20 Mb/s for replication traffic being consumed = 10 Mb/s) 10 Mb/s of available capacity or “bandwidth”.  Yes, 20 Mb/s of the 30 Mb/s is “consumed” by replication or whatever traffic you want to call it which means I have 10 Mb/s left so getting 355 Kb/s means there is a problem.  355 Kb/s isn’t anywhere near even 5% of the 10 Mb/s which is bad in anyone’s book.  In the derpgeineers defense we could be using the other 10 Mb/s or portions of it for other uploads say…to a backup cloud service but that wasn’t the case.  I also stated how that when iperf tests run from them to me, the bandwidth capacity was 27 Mb/s…perfect!  But, when I turn right around and run it from me to them…I get no better than 4 Mb/s.  I ran this test 3 times in a row from each side getting very similar results so it’s a replicable pattern! The fact I get such a stark difference from side to side is telling and should raise a red flag with anyone who knows these basic concepts!

It’s either a problem with you or with me.  I was actually telling them it’s us not you and we need to go back and look at our router…which is actually managed by another company so I can’t help…although we could…I can’t!  The derpgineer was getting defensive for no reason.  It’s all a bit frustrating not having our hands on all the pieces of infrastructure so it’s done correctly.  It’s like trying to drive where I hold the steering wheel, hey you other person…you push the gas while yet another person does the clutch and shifting…it’s all kind of a mess and anyone with common sense would see a car crash in our future!

Back to the issue.

Synchronous connectivity means you have the same amount of capacity / bandwidth both for your upload & download

Example:

30 Mb/s down, 30 Mb/s up

Asynchronous connectivity means you have a different amount of capacity / bandwidth depending on the direction.

Example:

250 Mb/s down, 30 Mb/s up

Most Internet these days is sold with a higher download speed than upload.  Charter, Comcast, ATT, Frontier, RoadRunner, Cox, etc, etc…they all say “Up to 100 meg down”, “lightning fast”, etc but they give you a 5 Mb/s upload which would be:  Asynchronous 100 Mb/s down, 5 Mb/s up.

Back to our “engineer” (derpgeineer) buddy.  I don’t know what he’s thinking because anyone who knows anything about anything can understand capacity.  Just like the 30/20 issue…I have 30 Mb/s and I’m only using 20 Mb/s so I have 10 Mb/s left of capacity.  It’s shocking and unbelievable that he doesn’t understand such a simple thing!  There’s not much of a difference either going to the Internet or over a VPN as the available bandwidth through the tunnel is the same minus a touch for encryption / decryption through the tunnel but that’s not going to consume almost 10 Mb/s of bandwidth (1/3rd of our capacity in this case).

Why am I talking about this?  Because even “engineers” seem to misunderstand that it’s not a problem with not ENOUGH bandwidth…it’s a problem of not having the ability for some reason TO USE the bandwidth you already have!

If someone ask you to upgrade your package you need to ask questions!  Why do I need to upgrade and supposing I do…what do I need to upgrade?  My upload, my download or both?  Sometimes you need more capacity (bandwidth) but sometimes you need to use what you already have better.  More bandwidth ALWAYS usually means more money out the door for not much benefit if you get more that you aren’t using efficiently.  I had a customer that upgraded to a gigabit download but only had a 10/100 Mb/s switch (gigabit = 1000 Mb/s).  There was no reason to get gigabit Internet UNLESS they also upgraded their switches to 10/100/1000!  Also…does your router do gigabit on the WAN link?  Many don’t!  Yowza…you can see why you need a team like ours.

If you’re an “engineer” and you don’t understand this simple concept of capacity and bandwidth then bone up and graduate from derpgineer to engineer and stop being a derp.

It’s a shame I even had to blog about this subject matter that’s how basic it is but it’s a problem many of the so called engineers just don’t understand.  It’s what a 1st year networking student learns…like; on the first day!

Update to this blog post (06/2019):

To further explain something that recently came up from a vendor I work with they explained to a customer that:

A “lower speed” dedicated fiber connection with low latency can carry as much data as a “higher speed” shared access best effort coax connection.

This is a 100% inaccurate statement and false information.

The key phrase here is “with low latency”, as though latency = bandwidth.  I like this post about the matter and they agree with my thoughts on the bandwidth issue where they state (link).

The relationship between latency and bandwidth determines whether we consider an Internet connection to be fast. Latency doesn’t affect bandwidth, but insufficient bandwidth can reduce latency. A high-latency, high-bandwidth connection would be slow to start downloading a Web page but would load quickly as soon as the download starts.

The other problem I have with the way they talk about fiber optic Internet is with the phrase about coax that says “best effort”, as though they’re trying less hard to transfer your information.

Remember, bandwidth is bandwidth is bandwidth. In no uncertain terms can a lower speed fiber connection carry as much data than a higher speed shared connection. That’s not how things work.

A “DIA” or “Dedicated Internet Connection” is a false term made up by god knows who in the world and calling coax “best effort” isn’t being honest because as we’ll see…everything is ultimately “best effort”. Nothing about the internet is “dedicated” because that’s not how it is built. It’s built on “peering relationships” and agreements from one ISP to pass the traffic of another ISP. Nothing is “dedicated” except between the carriers (ATT > Layer 3 > Comcast > Charter is but not your connection specifically) much like your driveway at your home (your customers Fiber connection). Sure, for a while you’re on a roadway all by yourself but then you hit the street. Then you might hit a two lane boulevard and then…the dreaded highway with traffic jams.  Outside of those “peering relationships” nothing is “dedicated”.  It’s a false notion that the connection at your location is “dedicated” other than your one physical path to some switch point that then links up to a trunk that connects however many customers.  The notion that you’re on some connection all by yourself is all just false.

So technically the “Internet” as we know it is a shared best effort connection. Nobody guarantees anything once you hit public infrastructure (the super highway road system of the Internet). To suggest you get some kind of guarantee from Fiber again is 100% false.

Don’t believe me?  Use “tracert” in Windows or “traceroute” on your Mac or Linux machine and see what you get.

Here’s what you get from a fiber connection:

traceroute microsoft.com
traceroute to microsoft.com (40.112.72.205), 30 hops max, 60 byte packets


3 hops I jump off from my ISP (not disclosing my hops)
* Pay attention to the bold underlined IP addresses

4 chi-8075.msn.net (208.115.136.27) 6.256 ms 6.241 ms 6.212 ms
5 ae31-0.icr02.ch2.ntwk.msn.net (104.44.237.21) 6.360 ms 6.337 ms 6.299 ms
6 be-102-0.ibr01.ch2.ntwk.msn.net (104.44.11.249) 87.031 ms 86.978 ms 87.078 ms
7 be-3-0.ibr01.was02.ntwk.msn.net (104.44.7.172) 87.679 ms 88.209 ms 88.118 ms
8 be-7-0.ibr01.nyc30.ntwk.msn.net (104.44.4.34) 87.110 ms 86.864 ms 87.474 ms
9 be-5-0.ibr01.ewr30.ntwk.msn.net (104.44.7.103) 87.248 ms 87.540 ms 87.372 ms
10 be-4-0.ibr01.sxl71.ntwk.msn.net (104.44.17.153) 87.182 ms 94.697 ms 87.225 ms
11 be-5-0.ibr01.dub07.ntwk.msn.net (104.44.16.113) 87.804 ms 87.815 ms be-7-0.ibr01.dub08.ntwk.msn.net (104.44.16.117) 95.596 ms
12 ae100-0.icr01.dub08.ntwk.msn.net (104.44.11.70) 87.633 ms ae100-0.icr01.dub07.ntwk.msn.net (104.44.11.60) 89.245 ms 87.978 ms

Here is what you get from a Comcast connection:

traceroute microsoft.com
traceroute to microsoft.com (40.76.4.15), 64 hops max

8 Hops and I jump off of the Charter connection I’m on
9 104.44.237.21 31.809ms 30.683ms 18.087ms
10 104.44.11.8 35.606ms 33.449ms 59.284ms
11 104.44.7.172 61.157ms 64.441ms 63.299ms
12 104.44.18.63 53.441ms 74.559ms 36.947ms
13 104.44.21.224 33.086ms 33.618ms 34.070ms

Fiber hops to 104.44.237.21 = 5
Cable hops to 104.44.237.21 = 9

Total amount of hops to the destination  that we can get to, because it stops at 104.44.21.224:

Fiber = 12
Cable = 13

Just because you have one more hop to the destination doesn’t make your bandwidth “faster”, it just makes your latency “less” but remember latency is NOT bandwidth.  Latency is just the time it taks for you to make a “call”, like “ping microsoft.com” until you get an “answer” or a ping response.

Just like that other post explains, once you establish the connection bandwidth is bandwidth is bandwidth, the download will come in and you will receive it at the advertised amount.  The only other caveat here is when the source doesn’t have as much upload as you do download.  This is another reason why the theory of “dedicated access” is all wrong.

If you have a 10 Mb/s coax, a 10 Mb/s fiber, and a 10 Mb/s copper connection THEN AND ONLY THEN would latency affect speed transfer in that it might take LONGER for the copper connection to initiate the transfer and if a lower latent connection like fiber or coax “started first” then it would finish first.  BUT; if you have a 10 Mb/s fiber, 15 Mb/s coax and a 20 Mb/s copper latency is moot because (depending on what you’re downloading) the other two mediums (coax and copper) will finish the download faster EVEN IF THEY TAKE LONGER TO ESTABLISH THE LINK! Why? Because to initiate the link / connection to the resource we’re talking MILLISECONDS!  Remember, the time in which it takes to make the call (for a resource) and the time it takes for the destination to respond (to answer back) is the latency.  Once the transfer starts your bandwidth is your bandwidth is your bandwidth.  The other two higher speed connections (coax and copper) will finish BEFORE fiber all day long, every time without question.

No, it’s not true that “a “lower speed” dedicated fiber connection with low latency can carry as much data as a “higher speed” shared access best effort coax connection.” <– NOT TRUE.  Alert, false, 100% “fake news” to use a parlance.  The higher speed connection will carry more data faster every time and it gets worse for the lower speed fiber the more data you throw at it like offline backups or downloads like Windows Updates which seem to be gigs and gigs of updates these days.

If you need assistance with anything technical including seriously simple to understand concepts other engineers might be having problems understanding like bandwidth and capacity (LOL) we’re the company you NEED to contact.

800-864-9497

Comments or questions are welcome.

* indicates required field

Leave a Reply

Your email address will not be published. Required fields are marked *