Sharing download bandwidth fairly AND efficiently... has to be a way?

[BCSteve]

But what you're describing... QoS... isn't what I'm looking for. My router does QoS pretty well, that wasn't the point. I want to share bandwidth equally, not just ensure higher priority packets get through during congestion. Perhaps I wasn't clear with my OP. I'm not as confused with QoS as you assume.

I already have it so, for example, VOIP or Skype doesn't crap out when Johnny's pulling DLC. That's not the issue. The issue is computer A is downloading at (nominally) 100% bandwidth and then computer B downloads at... well, some value that supposedly corresponds to the guaranteed upload that the QoS is giving it. Even if computer B has a higher priorty for the traffic type, computer A (defined as whichever was working first) always gets the lion's share. Again that can make sense (sort of) because QoS isn't there to ensure fairness... just to police minimums.

I want more.

I want computer A and B to both get (again, ignoring overhead) 50% while both are busy.

With the Asus, I *can* effectively limit A and B to 50%... which, by my logic, proves that download can be managed. Its just not smart enough to do that dynamically. It shouldn't be that hard. The router should just know that B isn't using its 50% and then bump A's up to 100% until B starts to make demands.

That's not QoS... I get that. I guess there isn't a name for what I'm looking for if we don't want to call it QoS. It isn't really traffic shaping either because I'm not trying to prioritize types of traffic. I'm actually trying to do something simpler, I think, and that's ensure each [arbitrarily defined] party gets what they paid for. At a minimum, though, not at a constant.

It does appear that Router OS can do this. I'm still not convinced Ubiquity can, after a lengthy back-and-forth with their support folks. In the end, they said it was an "interesting idea that may be looked at in the future". Is it really that unusual? To me, its obvious... in any case where an Internet connection is shared between multiple unrelated networks, and in some cases with related networks or devices on a network.

 

[BCSteve]

HFSC (man I hate random acronyms) seems to describe what I want. What I'm unclear on... how to implement that?

Google tells me that HFSC is built-into Linux (that's awfully vague). I don't really want to build a Ubuntu box or whatever just for a router.

 

[BCSteve]

I found a wired-only RB2011 for $85 with free shipping. I can't see how I can go wrong with that, even if it just ends up being a learning exercise. I think I might pick that up along with their lowest-end hEX lite product because I'm curious how the same OS can run on such vastly different hardware and I want to see if my real world application actually needs the extra power and features. If I find out the RB2011 can't handle what I need (which would surprise me) then so be it... learning experience.

 

[sfx2000]

But what you're describing... QoS... isn't what I'm looking for. My router does QoS pretty well, that wasn't the point. I want to share bandwidth equally, not just ensure higher priority packets get through during congestion. Perhaps I wasn't clear with my OP. I'm not as confused with QoS as you assume.

I already have it so, for example, VOIP or Skype doesn't crap out when Johnny's pulling DLC. That's not the issue. The issue is computer A is downloading at (nominally) 100% bandwidth and then computer B downloads at... well, some value that supposedly corresponds to the guaranteed upload that the QoS is giving it. Even if computer B has a higher priorty for the traffic type, computer A (defined as whichever was working first) always gets the lion's share. Again that can make sense (sort of) because QoS isn't there to ensure fairness... just to police minimums.

I want more.

I want computer A and B to both get (again, ignoring overhead) 50% while both are busy.

You are confused - can't have more... can over-commit, and that's ok, but once you get to the limit, the queue's are what they are... again, QoS is reservations, not limitations - most consumer grade routers these days, if you let them do QoS, they do an ok job - but it's not as simple as you think...

See here -[QoS] - Quality of Service in a nutshell

 

[BCSteve]

Why would I "see here"? Again, I'm not confused on what QoS is. I'm quite clear on the fact QoS isn't what I'm looking for. More like "smart limiting" in conjunction with QoS.

 

[kvic]

You don't need a Mikrotik or another new router to meet your requirement. Maybe they give you a GUI to finish the same job...or come with configuration templates making life a bit easier..

To DIY, it does require somewhat in-dept understanding of QoS in Linux which you seem to possess...so shall be a good start from here:

http://wiki.mikrotik.com/wiki/Manual:HTB

Quoting the Mikrotik's link because I think that page illustrates well for your case.

Now you can achieve the same on RT-N56 with 'tc' (traffic control) ...finally put it into a script..and share with the community here.

 

[BCSteve]

Wait a sec... are you about to blow my mind? Are you saying I can script my RT-N56???? I'm off to Google.

Edit: Damn! I had no idea I could Telnet to this box. I'm not sure what I can do with that... but I'm in, so time to explore.

 

[L&LD]

Perhaps. I don't really think so from my understanding of their "Adaptive QoS" it just *sounds like it should* do what I want, but in effect I *think* it isn't anything more than a simpler QoS interface. I could be wrong... that's not the router I have. I've not been very happy with Asus products, so I'm not too inclined to jump back on that ship.

Basing your conclusions on a five year old product that was never considered 'great' is not a fair way (to yourself) to judge any current model (and the RT-AC56U is a whole different beast than what you have).

You are also assuming (wrongly, I feel) about how ineffective Adaptive QoS is too.

For between $50 to $100 you can try this model (depending on your location and any 'sales' you can wait for) and see if this works or not.

The direction you're currently going down in this thread seems painful to me to try to re-invent the wheel. In time, money and hardware. Not that you won't get to where you want to be eventually, but I don't see anywhere in your posts so far that needs that kind of heavy-handedness to achieve your end goal.

 

[BCSteve]

You could be right. My conclusion isn't based on just my router... I said "not been very happy with Asus products" and that goes beyond my router. Frankly, I've never understood why they enjoy such a reputation. Two motherboards, an O!Play video player, a monitor, that ill-conceived "Transformer Prime" (how they didn't get sued by Matel...) and this router. Asus just rushes to put out the next thing and doesn't bother to support their last thing. Their focus is trend, not reliability. For a router, I want reliability. Is the AC56U that much better? It very well may be, but as I said.. I'm not inclined to support that company based on my experiences thus far. At the time I bought my N56U it was easily the top-rated router on Amazon, Newegg, etc and yes it was considered 'great'... it absolutely hammered the competition on the lists here at snb at the time. It wasn't cheap either as I believe I dropped almost $250 on it.

But you are very well possibly right that I'm overlooking simpler solutions. I dug up an old DIR-615 I had in the closet after seeing it was the only thing I have capable of (barely) running DD-WRT. I'm going to test that today to see if it can do what I want (the settings suggest it is, but the proof will be in the tests). It isn't the solution... the hardware just isn't up to the task of my network... but it'll prove to me whether or not that firmware has the capability. I've somewhat ignored DD-WRT because it is - by its nature - focused on wireless. We'll see

 

[L&LD]

You also may be right that an Asus router won't give you want you need. But side testing other firmware and hardware and then assuming that modern firmware and hardware are similar is a major mistake you're making, imo.

Adaptive QoS has worked for me and various customers where Tradition QoS (Asus' terms) did not.

I don't think it matters how much you may understand how QoS works. The implementation is what counts (and I don't think anyone knows what Asus is doing between their methods).

All I'm saying is do a valid test on the actual equipment that supports it.

High ratings on Amazon, Newegg, etc. mean very little (the people 'reporting' usually have little networking knowledge and certainly less standards testing that they offer for their ratings) and the once high cost of a product also isn't an indication of it's worthiness today.

The original ENIAC is worth the equivalent of over $6M today. Even if I had the room, electrical requirements and the know how to use it, it is still worthless to me and almost everyone else.

When I started buying computers for myself, the price range was in the $3,000 to $4,000 range for a 'good' one. Today, that is still true. The difference is that the equipment today is 3000x better and effectively almost 4 times cheaper (even if the absolute amount is the same).

Don't get stuck on what was. Take the time to acquaint yourself with what is current, now.

 

[BCSteve]

That's a lot of random advice, but I don't see the application. Telling me about the economics of past technology is for what? Did you somehow think I ever said that my router was awesome? If it was awesome, I wouldn't be here. I'm not arguing your points... they are all given. The sky is also blue and gravity works but none of that has to do with my problem.

Your most salient point - "I don't think anyone knows what Asus is doing between their methods" - is exactly the reason I shouldn't be buying a newer Asus in hopes it maybe does what I want, based solely on the fact the name of it ("Adaptive") kinda sounds like something that might do the trick. I did look into it. Their own sparse and vague explanation of what "Adaptive QoS" is suggests its really nothing more than normal QoS without the need to set anything. NOTHING about it (other than the name) suggests is dynamically adjusts ingress limits to fairly share bandwidth. I see no point in throwing more money at a company I don't trust when there are less expensive options that have a far better chance of getting me where I need to be (Mikrotik or maybe DD-WRT).

But thanks anyway.

 

[L&LD]

That's a lot of random advice, but I don't see the application. Telling me about the economics of past technology is for what? Did you somehow think I ever said that my router was awesome? If it was awesome, I wouldn't be here. I'm not arguing your points... they are all given. The sky is also blue and gravity works but none of that has to do with my problem.

Your most salient point - "I don't think anyone knows what Asus is doing between their methods" - is exactly the reason I shouldn't be buying a newer Asus in hopes it maybe does what I want, based solely on the fact the name of it ("Adaptive") kinda sounds like something that might do the trick. I did look into it. Their own sparse and vague explanation of what "Adaptive QoS" is suggests its really nothing more than normal QoS without the need to set anything. NOTHING about it (other than the name) suggests is dynamically adjusts ingress limits to fairly share bandwidth. I see no point in throwing more money at a company I don't trust when there are less expensive options that have a far better chance of getting me where I need to be (Mikrotik or maybe DD-WRT).

But thanks anyway.

I cannot make you understand my points, that is up to you. You have managed to mangle my intent though.

Are you not able to buy and return if it doesn't work? You don't need to throw money at Asus without some kind of recourse available to you (even if you just resell the router privately afterwards).

I am stating what I have because Adaptive QoS has worked for many here, including myself and some customers too where Traditional QoS didn't, previously.

Your conclusions based on reading marketing materials about it are, uhm, baffling.

I've tried to help as well as I can. I'll be reading and hope to see how you solved this issue soon.

 

[BCSteve]

Baffling? Baffling is suggesting that because Adaptive QoS "has worked for many here, including myself" that it will somehow work for a non-QoS issue.

Had you told me (and you have not) that you understood my initial needs and you have experience suggesting that Adapative QoS does, in fact, resolve the issue at hand... that would be different. But really what you said is more akin to this: "Johny's car needs an oil change or it won't work. I like oranges. Johnny should get an orange."

I have no idea at all what you're referring to with "marketing materials". And no, I don't agree with throwing money at something (particularly to a company I don't wish to support) in hopes it works... that's why we do research, no?

 

[L&LD]

Baffling? Baffling is suggesting that because Adaptive QoS "has worked for many here, including myself" that it will somehow work for a non-QoS issue.

Had you told me (and you have not) that you understood my initial needs and you have experience suggesting that Adapative QoS does, in fact, resolve the issue at hand... that would be different. But really what you said is more akin to this: "Johny's car needs an oil change or it won't work. I like oranges. Johnny should get an orange."

I have no idea at all what you're referring to with "marketing materials". And no, I don't agree with throwing money at something (particularly to a company I don't wish to support) in hopes it works... that's why we do research, no?

Okay, I give up. You know more than me. Good job.

 

[ml70]

I want computer A and B to both get (again, ignoring overhead) 50% while both are busy.

You need to get down and dirty with tc. What you want to do is to create a leaf for each ip address in your subnet, as opposed to the way stock Asus does it with just one main leaf and 5 classes.

So you'd make a shell script which creates leaves for ip's 0-255 and each of these will have 5 classes, or more. Not sure if Asus has enough ram or cpu power for that.

If you only have a few groups that need managing it's easier, let's say computers A, B and the rest in C, you just create separate leaves for each 3, that's well withing the capabilities of I'd say any Asus router that comes with qos.

You can mitigate the separatedness of these groups by using a properly planned 3 (or more) band pfifo_fast for eth0, under which the qdisc leaves are. So that certain packets from all groups get final preference in the last queue before metal.

After figuring out all this, I personally decided not to go forward with separate ip leaf creation and simply configured Asus' stock qos in a way that provides a somewhat satisfactory experience - find out what is really consuming most bandwidth and can it be prioritized down.

Google is notorious for not playing very nicely along with bandwidth limits, you have to dig out all google server ip ranges if you want to limit youtube, android updates, googleupdate on windows etc crap which will make internet very painful for everyone else.

 

[System Error Message]

mikrotik is much easier than writing scripts and trying to do layer 3 routing with a consumer router.

The thing about various sites is while they do have multiple IP addresses they do also use different protocols for updates and such. What you need to do is find out what port and protocols are used for updates and incorporate that. What isnt directly clear with mikrotik is that you can type a protocol number instead of the ones in the list. So if there isnt any information of what port or protocol is used than using wireshark or even mikrotik's own sniffer can help determine it but it is quite a lot of work trying to find the right packet.

After you have done your research you need to learn to divide your QoS into categories and from there into sub categories and so on. One thing people tend to forget with configuring a router regardless of brand is that they tend to forget to account for everything else. You would also need to define how you want to handle the uncategorised traffic in terms of QoS by assigning a priority. While mikrotik routerOS lets you change the QoS type and such make sure you read their guides and wiki because each algorithm has benefits and the use of buffers can create latency. However certain things dont need to worry about latency such as streaming or downloading but you have to be careful what you set or you could make things worse.

 

[BCSteve]

I have a Mikrotik 2011 on order... I'm looking forward to playing with it. Their official guides are pretty sparse, but there are plenty of resources out there. I have a good feeling I'll be able to achieve what I want to achieve with this machine. Thanks again to abailey for throwing that name out there. Right now I'm thinking I wish I would have heard of them years ago. Interesting that their stuff isn't even on SNB's router charts at all (Ubiquiti is, which is the only reason I ever heard of them.)