Re: new thread: QoS, latency, bandwidth and the FCC/net neu…

Top Page
Attachments:
Message as email
+ (text/plain)
+ (text/html)
+ (text/plain)
Delete this message
Reply to this message
Author: Herminio Hernandez, Jr.
Date:  
To: Main PLUG discussion list
Subject: Re: new thread: QoS, latency, bandwidth and the FCC/net neutrality debate
Let me further explain what I "what these people know". They are people who
pushing claim that packets are just packets who know very well that is
simply not true. This is not a debatable point. I can show you packet
traces of a http connection, a https session, a ssh session, and a
streaming session. Just looking at how they transmit without even diving
into the details will demonstrate how silly that idea is. Those are just
four types! We have not talked about torrents, XMPP, SMTP, DNS, POP, BGP,
SIP, FTP, Telnet ... I can go on but none of those packets behave the same
way and to even think that there people who 'honestly' think that we can
one policy to manage it all is beyond belief. That is what bothers people
(I mean those at the top of this whole debate) do know better.

On Wed, Nov 29, 2017 at 12:20 PM, Herminio Hernandez Jr. <
> wrote:

> There is avoiding because of supply and demand. The internet has exploded
> in bandwidth use. Even if all the ISPs had 100Gbps backbone connections
> there will come a point when even that is not enough. Plus it is not just
> port speed. Even if you had 1 Tbps (which I am pretty sure does not exist)
> ports you need a CPU fast enough to be able to process the frames and
> packets at line rate. You need enough memory for support the buffering.
> None of this is cheap. So business made the decision it is cheaper to
> oversell at a cheaper price. Now you can “demand” that ISPs must always
> upgrade. However you will not like your bill and that will pretty much
> guarantee no new ISP startups since they probably will not have the capital
> for top line Network gear.
>
> Sent from my iPhone
>
> On Nov 29, 2017, at 11:36 AM, Carruth, Rusty <>
> wrote:
>
> No, sorry, your understanding of “what these people know” is not correct.
> I do not believe that we should let those who ‘hook stuff together’ hide
> the fact that they have been overselling bandwidth for a long time.
>
>
>
> Probably from the very beginning.
>
>
>
> In the beginning (well, not really, but in 1994-ish time frame), if you
> had a T1 line, you could tell who had a 56K connection, because an FTP or
> other file download from them would max out at 56k. (I know because I was
> there and did that. Many times. Sometimes I’d see it bounce around just at
> and below 56K as others used the wire over which I was transferring data (I
> usually assumed it was on their end, but don’t really know for sure)).
>
>
>
> In those days, there were fewer end points with fewer users and no or very
> little ‘realtime’ data, so ISPs could get away with selling more bandwidth
> than they had, or perhaps more reasonable would be to say that the formula
> they used (if any!) for computing how much bandwidth they needed between
> themselves and ‘everyone else’ assumed no realtime data (we didn’t really
> have much if any ‘realtime’ data back in those days as far as I can
> remember), and less loading per customer (that is to say, most customers
> would not utilize the full bandwidth (or a significant portion) of their
> pipe for HOURS at a time).
>
>
>
> Now that the customer has changed, it is time for the ISPs to realize that
> they need to change the formula. QoS will only postpone the inevitable
> (and probably not for long!). I’m pretty sure there is not enough
> inter-ISP bandwidth to rationally serve the ‘needs’ of their customers, for
> much longer.
>
>
>
> So, I disagree with the conclusion that ‘there is no avoiding it’ (4th
> sentence below). What there is no avoiding is the fact that ISPs don’t
> have enough ‘backbone’ bandwidth to handle all their customers. QoS won’t
> fix that, it’s just a bandaid to reduce the available bandwidth for
> non-realtime data users until there is not enough bandwidth even for the
> realtime folks. (I could have said ‘steal bandwidth from normal,
> non-realtime users’ if I wanted to be more pejorative and say what I really
> feel ;-).
>
>
>
> And, just like running out of IPv4 addresses – that time WILL come. With
> or without QoS. (And unfortunately NAT won’t fix THAT problem ;-)
>
>
>
> Your guess as to when that will be is as good as or better than mine…
>
>
>
> Will the day come that ISPs become responsible to their customers to
> actually provide what they are advertising? I doubt it. Read your fine
> print.
>
>
>
> (Now, if we had the infrastructure I was advocating for a while, there
> might actually BE enough bandwidth on the backbone. Oh, well, that will
> probably never happen)
>
>
>
> Anyway, if the ISPs didn’t grossly oversell their available bandwidth,
> this whole debate would sound completely silly.
>
>
>
>
>
> Rusty
>
>
>
> *From:* PLUG-discuss [mailto:plug-discuss-bounces@lists.phxlinux.org
> <>] *On Behalf Of *Herminio
> Hernandez Jr.
> *Sent:* Wednesday, November 29, 2017 10:56 AM
> *To:* Main PLUG discussion list
> *Subject:* Re: new thread: QoS, latency, bandwidth and the FCC/net
> neutrality debate
>
>
>
> Even if there was no voice and video QoS would be needed for simple fact
> that network congestion exists. QoS has been around since the days of
> dialup. The world where there is unlimited bandwidth and network interfaces
> will always transfer at line rate simply does not exist. There is no
> avoiding it. People will always want more data faster than new
> infrastructure can be built.
>
>
>
> This what I find disturbing about the whole debate. There are people who
> are pushing the idea that packets are just packets and voice, video, data
> can all be treated the same. These people know better. They know it is not
> a question if we should manage the internet but who will manage it. Who’s
> traffic will get priority service, who will not?
>
> Sent from my iPhone
>
>
> On Nov 29, 2017, at 10:07 AM, Carruth, Rusty <>
> wrote:
>
> I strongly disagree with the statement “which the internet needs to
> function”.
>
>
>
> No, the internet does NOT need QoS in order to function. Its been working
> fine for years without that. Its just people trying to do things on the
> internet that it was not designed for who demand QoS in order to co-opt the
> internet for THEIR use.
>
>
>
> If you insist that the internet MUST have QoS to function, then that’s the
> end of the discussion. Those who believe that must demand NO NN, otherwise
> the internet won’t work the way they think it should. Those who have not
> bought in to that assumption may be on either side of the debate. But if
> you buy the theory that QoS is required for the internet to function then
> you must oppose anything that allows the internet to function the way it
> was designed.
>
>
>
> And the point about QoS effectively stealing bandwidth from other users is
> something we’ve not spoken of thus far, as far as I can remember. But it
> is something to keep in mind – hacking the medium to enable realtime data
> reduces the usability of the internet for all people who are not using
> realtime data.
>
>
>
> Which brings up a rabbit trail which I’ll start a new thread upon.
>
>
>
> Rusty
>
>
>
> *From:* PLUG-discuss [mailto:plug-discuss-bounces@lists.phxlinux.org
> <>] *On Behalf Of *Herminio
> Hernandez, Jr.
> *Sent:* Tuesday, November 28, 2017 3:33 PM
> *To:* Main PLUG discussion list
> *Subject:* Re: new thread: QoS, latency, bandwidth and the FCC/net
> neutrality debate
>
>
>
> Here is a good definition of QoS from Cisco: "The ability of the network
> to provide better or 'special' service to a set of users/applications to
> the detriment of other users/application". Net Neutrality cannot exist in a
> network where QoS is needed which the internet needs to function.
>
>
>
> On Tue, Nov 28, 2017 at 5:15 PM, Herminio Hernandez, Jr. <
> > wrote:
>
> I understand your frustration, but to be frank it is unrealistic to think
> that the industry is going to redesign the physical infrastructure to
> accommodate voice and video. The ship has sailed there. Converged
> infrastructure is here to stay. Now the job is to find the best solution
> for this reality and Net Neutrality is not it IMO.
>
>
>
> On Tue, Nov 28, 2017 at 5:05 PM, Carruth, Rusty <>
> wrote:
>
> I’m going to have to switch to inline answers. See below.
>
>
>
>
>
> *From:* PLUG-discuss [mailto:plug-discuss-bounces@lists.phxlinux.org] *On
> Behalf Of *Herminio Hernandez, Jr.
>
> >TCP would not solve the issue. Think about constantly having to ask the
> person on the other end of a phone conversation to repeat themselves
> because the sound kept dropping. That would drive you be insane. That is
> very much like TCP. >Voice and Video traffic simply will not work in that
> scenario.
>
>
>
> Which is pretty much to my point. TCP doesn’t work well for realtime data
> (unless perhaps you have nobody else on the wire and a perfect wire).
>
>
>
> So, the first attempt at a workaround was to use UDP, whose performance
> fits better with ‘almost realtime’ data in a network that was fairly
> quiet. When that began to fail because of busy networks, something else
> was needed.
>
>
>
> The next attempt seems to be to change the network transport protocol to
> prioritize certain packets over other packets, which is IMHO risky business.
>
>
>
> IF, and ONLY IF, there is absolutely no allowance for a transporter of
> packets to give (or remove) special priority to certain packets based upon
> something other than their type (VoIP, video), then the issue of realtime
> data on the interent MIGHT have found a way out of the problem of trying to
> force something onto a medium which it wasn’t designed to handle. But I
> still feel this is trying to force a design onto something that can’t
> handle it.
>
>
>
> In any case, I still think that those who use ‘the internet’ for realtime
> data and wish to force it to do what it was never designed for have MUCH
> more of a requirement to ‘play nice’ than those who use it for what it was
> originally designed.
>
>
>
> > You are right ethernet was not designed for voice and video in mind,
> but that is where we are at and it is not changing.
>
>
>
> So then you should reject any attempt to cram a bad design onto something
> that wasn’t designed for it. Which those against any sort of net
> neutrality seem to be trying to do – force a bad design on the wrong medium
> (assuming I have half a clue as to what NN is SUPPOSED to be).
>
>
>
> Those who wish to transport realtime data over a network should design a
> network that can do that, not co-opt somebody else’s network. Again, IMHO.
>
>
>
> Rusty
>
>
>
> > On Tue, Nov 28, 2017 at 4:37 PM, Carruth, Rusty <
> > wrote:
>
> I still disagree.
>
>
>
> First, if they needed reliable delivery of packets, then they should use
> TCP.
>
>
>
> My understanding of the ‘theory’ of why streaming services use UDP is that
> it doesn’t hurt ‘much’ if you lose a ‘few’ packets – not as much as them
> showing up in the wrong order, or massively delayed, so using UDP is a
> workaround to try to use a medium that wasn’t actually designed to carry
> realtime data.
>
>
>
> So, I go with the line of reasoning that claims that using ‘the internet’
> for real-time data is to misuse the medium. And if a medium is misused,
> those so misusing it shouldn’t be surprised if it doesn’t work in a way it
> wasn’t designed to do.
>
>
>
> Yes, it doesn’t work well with real-time data.
>
>
>
> Wasn’t intended to, IMHO.
>
>
>
>
>
> (Just a grumpy old man who knows that the internet pre-existed the guy who
> claims to have invented it… And who even knows what ftp, telnet, rcp,
> gopher, and uucp used to mean ;-) (and who performed tests to prove that,
> between two Solaris boxes on a COAX ‘ethernet’ cable, FTP was faster than
> anything else. But I digress! ;-)
>
>
>
> *From:* PLUG-discuss [mailto:plug-discuss-bounces@lists.phxlinux.org] *On
> Behalf Of *Herminio Hernandez, Jr.
> *Sent:* Tuesday, November 28, 2017 2:28 PM
>
>
> *To:* Main PLUG discussion list
> *Subject:* Re: new thread: QoS, latency, bandwidth and the FCC/net
> neutrality debate
>
>
>
> Rusty,
>
>
>
> I know my language was strong but let explain why, First not all traffic
> behaves the same. Go back to my initial post on the differences between TCP
> and UDP. UDP by the nature of the protocol is more sensitive to things like
> packet loss, latency, etc. So in order to deliver UDP services reliably (ie
> most streaming services) some type of prioritization must occur. If not
> then video will be constantly buffering and VoIP calls will drop. The
> reason why there exist QoS policies is because engineers are try to work
> with the transport medium we have. Bandwidth is a limited resource and you
> have all these different types of traffic contending for the same resource.
> If people expect web browsing, YouTube, Mumble, Netflix, SFTP, all run
> efficiently across the wire then prioritization is a reality that will not
> go away. This is nature of modern networks where data, voice and video are
> all converged on the same media. The reason I used the language I did was
> b/c an engineer who does not understands this and actually thinks that 'all
> traffic' can be treated the same will actually bring harm to the network.
> He will be doing a great disservice to users he supporting all under the
> false notion of 'equality'.
>
>
>
> On Tue, Nov 28, 2017 at 2:38 PM, Carruth, Rusty <>
> wrote:
>
> Yes, lets get back to the technical issues.
>
> First, though let me review: Apparently an ISP has been targeting certain
> SITES or DOMAINS and throttling them. If that the case, then a discussion
> of the network issues is beside the point - the issue of treating certain
> endpoints differently based upon some non-technical issue would be the
> issue.
>
> Anyway, that being said -
>
> I was actually somewhat offended when the statement was made claiming that
> anyone who believes that all traffic, regardless of type (voice, file, web
> pages, etc) should be treated the same was an idiot.
>
> On what basis is someone who thinks that a certain type of traffic
> DESERVES a different assurance of throughput against any OTHER type of
> traffic? If the entity using a certain transport mechanism has different
> requirements than the transport medium can provide, then they are the
> unwise ones. And have no right to demand that the transport medium change
> to accommodate their demands.
>
> Especially at everyone else's expense.
>
> Why does VoIP or Video REQUIRE special treatment? I claim that either the
> systems which use these technologies either figure out ways to work within
> the limitations of the medium, or find a different medium. Don’t demand
> that the medium ADD special treatment for you.
>
> One might then say that having the user pay extra for the special
> treatment would address this, and not force the cost of this on to all
> users, but this opens the door for a medium provider to use their
> (essentially) monopoly position to materially affect the open market in
> ways which could easily damage the open market.
>
>
> (I was tempted to say something about 'in the beginning, all traffic was
> just packets - and they still are just packets'. ;-)
>
> All the above has NOTHING WHATSOEVER to do with the company I work for,
> its IMHO.
>
>
>
> -----Original Message-----
> From: PLUG-discuss [mailto:plug-discuss-bounces@lists.phxlinux.org] On
> Behalf Of Herminio Hernandez Jr.
> Sent: Tuesday, November 28, 2017 7:44 AM
> To: Main PLUG discussion list
> Subject: Re: new thread: QoS, latency, bandwidth and the FCC/net
> neutrality debate
>
> I do not what you are getting at? Yes we all look at Net Neutrality
> through the lens of our assumptions on how the economy should be built. I
> am sure many would believe that government should a significant role is
> managing and others not. Most of this thread has focused on that.
>
> I would love to discuss more the technical side of the debate. The first
> part of original post thread were the technical reasons why I felt NN was
> bad policy.
>
> Sent from my iPhone
>
> > On Nov 28, 2017, at 7:24 AM, Steve Litt <>
> wrote:
> >
> > On Mon, 27 Nov 2017 22:52:04 -0700
> > "Herminio Hernandez Jr. " <> wrote:
> >
> >> First since I do not believe in
> >
> >> central planning
> > ^^^^^^^^^^^^^^^^
> >
> >> I do not know what
> >> competitors will once they have the freedom to offer services. This
> >> what is awesome about the
> >
> >
> >> Free Market,
> > ^^^^^^^^^^^
> >
> >> if there is market that was
> >> moved closed off now open they will find creative ways to provide
> >> services.
> >
> > Looks to me like Net Neutrality is being used as a proxy for some
> > much more generic theories.
> >
> > SteveT
> >
> > Steve Litt
> > November 2017 featured book: Troubleshooting: Just the Facts
> > http://www.troubleshooters.com/tjust
> > ---------------------------------------------------
> > PLUG-discuss mailing list -
> > To subscribe, unsubscribe, or to change your mail settings:
> > http://lists.phxlinux.org/mailman/listinfo/plug-discuss
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
>
>
>
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
>
>
>
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
>
>
>
>
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
> ---------------------------------------------------
> PLUG-discuss mailing list -
> To subscribe, unsubscribe, or to change your mail settings:
> http://lists.phxlinux.org/mailman/listinfo/plug-discuss
>
>

---------------------------------------------------
PLUG-discuss mailing list -
To subscribe, unsubscribe, or to change your mail settings:
http://lists.phxlinux.org/mailman/listinfo/plug-discuss