Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[icemankent]

Today I decided to give this router a try.
It came with firmware version 1.04 (hardware A1) on it out of the box.

I configured it for the most part, identical to my dgl-4500 regarding clients, IPs, etc.

I noticed one anomaly regarding the wireless setup.
For the 5 Ghz band, I chose 802.11n only.
Even with that setting, it chose to communicate with my Samsung Galaxy S4 as 802.11ac.
I verified this on the wireless status page.
In fact, there was one client using N and one using AC - even though I specifically configured the router to be N-only.

My REAL beef, however, is with the broken QoS as it relates to the DOWNlink speed.
- With QoS turned OFF, I get 3Mup/35Mdown (speedtest.net)
- With QoS turned ON, I get different results based on the queue level - even if there is no other traffic in any other queue, and/or no other clients connected

I tested this up/down/sideways and still got the same results which I will detail below.
Uplink speed 3M, Downlink speed 35M.
I set up the QoS as "strict priority queue" - for simplicity sake.
Now, QoS is DESIGNED to work such that its rules ONLY get enforced in times of congestion.
Clearly this is not what is happening.
It would seem that DLINK has chosen to LIMIT the maximum bandwidth based on the Queue you choose.  This is madness.
If there is no other traffic, any client should have access to the full bandwidth available.
This is a HUGE deal-breaker and DLink should be ashamed of themselves for being in this business and not knowing how QoS is supposed to function.

SETUP:
On my WIRED pc, I gave it a priority of [4] Best Effort (default).
On one laptop (next to the router, connected on 802.11n), I gave it a priority of [2] Higher.
On another laptop (next to the router, connected on 802.11g), I gave it a priority of [3] Normal.

I tested my WIRED pc (priority 4) on speedtest.net (note, NO traffic on either of the two wireless laptops).
I could easily get 3M up, but ONLY 20M down.

I thought this was strange, so I tried my laptop (the one set to priority 2).
Speedtest.net gave me 3M up, and around 32M down.

I wondered what was going on so I tried the second laptop (the one set to priority 3).
Speedtest.net gave me 3M up, and around 28M down.

What this router is doing, is looking at the priority level associated to the queue and LIMITING its maximum bandwidth - even when there is NO OTHER TRAFFIC competing with it !!!

To test this further, I gave my WIRED pc the highest priority [1].
Speedtest.net now gave me 3M up and around 33M down.
This is still not my maximum possible downstream speed (35M) - even with no other traffic, and selecting the absolute best queue !!

So, at this point, I turned QoS OFF.
On the same WIRED pc, now I was getting 3M up and 34.9M down !!!

DLink, your QoS is shameful.

How is it that you got your QoS SO RIGHT on the dgl-4500, and on your latest models you completely missed the boat regarding QoS functionality ?

There's no way I can continue to own this router if simply enabling QoS negatively affects my download speeds - and setting up actual Queues makes it even WORSE.
Nope, this black garbage can is going back to the store.

Someday DLink may get it right, but after reading how they botched the new DGL-5500 and seeing how this one behaves, I'm not holding my breath.

[FurryNutz]

Funny, I get 50/3 when using any of my routers including the 868L. I just helped my friend set up 2 of these 868Ls, one in router mode and the other in bridge mode for gaming. He's getting 60/5 speeds from his ISP via his SB 6580 modem with the 868L in the DMZ. Game consoles are connected to the 2nd 868L in bridge mode and working on 5ghz AC mode. TS is set for 100/10. QoS is strict and set up for Normal for general devices and Highest for game consoles.

I think your attempting to apply DGL logic and operation to a DIR series router which is completely different.

I'll check to see of setting single N mode on 5Ghz results in AC still being used. I wouldn't think thats a major issue though. I'll test and verify.

Good Luck.

[icemankent]

Thanks for looking into this on your side FN, but the bottom line of all of this is that using QoS degrades the speed available to devices.

QoS OFF - all is good - no performance "hit" to speeds
QoS ON - issues with speed of varying degrees - directly proportional to the QoS setting.

Also, it [only] affects the downstream speed, not the upstream.

[FurryNutz]

I don't see any hit on the downside at all on mine. I'll have another look however I have done extensive testing for D-Link and results are great for this router. Both Isolated and online. Not sure why your seeing a performance degrade no yours.

[Doctor Doom]

I would like to add a bit of food for thought.

In my experience with "QoS", one frustrating thing is QoS isn't exactly standardized. Unless you are talking about 802.1p but that's a layer 2 implementation vs layer 3. I've played with many routers from different vendors, generations of vendors as well as corporate firewalls with QoS and they all behave differently with the one common goal of trying to manage traffic either by priority and/or traffic shaping.

QoS isn't necessarily designed to work in such that it's rules ONLY get enforced in times of cogestion. There isn't a consortium that decides that a router is "QoS V3" certified. In some implementations a QoS engines will only really kick in if the network is saturated.

Between the 868L and 4500, I don't know which would truly perform better under high traffic loads in real world conditions. The 4500 uses gamefuel and only concerns upstream traffic only. The purpose was strictly to ensure a quality of service from a lag/ping perspective only. If someone uploading via FTP while someone is playing battlefield, the pings would remain resonably low for the gamer vs having QoS disabled or a non QoS router.  The 868L can define both up and down with strict queues "pipes". 

QoS was a bigger play years ago when bandwidth was at a premium. Those with 5/1 DSL connections would see a difference in terms of lag. Now we see 50/10 and even higher and even in a household with 2-3 skype connections and someone playing online games you probably won't be able to tell the difference. Now if someone is saturating the network because of say bit torrent, it's MUCH easier to limit the upstream from the application rather than tinkering around with QoS settings on a router.

If you are looking for a killer QoS engine, the DGL-5500 is worth investigating. It handles everything from priority and dynamic traffic shaping based on application. Not only would it be able to detect if traffic is youtube, it will even know the resolution and will change the traffic accordingly.

[icemankent]

Thanks for your feedback.
It is entirely possible that I got a lemon.
Unfortunately I don't have the patience to go buy another one to try out.
Well.....maybe in the future, perhaps.

When I see some (good) feedback on the [firmware] of the newest Asus rt-ac68u, I may give that one a try.

I set up many different enterprise-class routers with QoS (Cisco, Nortel, etc), and the QoS works as it is supposed to.
The bottom line is that QoS never (needs) to kick in until congestion occurs.
When there is congestion, the QoS rules then start to shape the traffic based on the priorities you assigned to the various queues.
This is simply QoS 101.

[FurryNutz]

Thank you for your feedback and experiences.

Good Luck.

[FurryNutz]

Just FYI, testing QoS in this manor may not quality this as a complete valid evaluation of the performance of QoS when only testing with 1 PC. QoS really kicks in with several devices on line and connecting to various service so that QoS can do its thing. Testing with only one device doesn't quite do anything for QoS since it's the only device online. What does QoS need to do if only 1 device is online and nothing else? QoS is to ensure that more than one devices gets good data flow while other devices are online and ensures they get good service as well.

The DIR-868Ls QoS does a fantastic job with what I have experienced and tested. I really like the 868L and it's performance.

People shouldn't be quick to judge before they experience and review ALL the facts!!!

[icemankent]

Agreed, however even with 1 pc being used at a time, why on earth would simply turning QoS ON, end up negatively affecting the speed of the downstream data transfer ?

It may very well be that the router I bought was a lemon, because simple logic would say that if there's only 1 device - no matter what - turning on QoS should not degrade the performance whatsoever.

I might buy another one in the coming weeks and test it out.
If I do so, I'll report back here what comes of it.

[FurryNutz]

Well, not knowing of what else you are connecting, testing and of your environment and conditions, I'd say there is a slight chance of there is something in your test evaluation that could be impacting your result and possibly creating a false negative. Many variables can cause problems.

[FurryNutz]

Working on getting GPL code for rev As. Until then:
http://forums.dlink.com/index.php?topic=53989.msg270106#msg270106

[FurryNutz]

Can anyone with 100Mb or greater down ISP speeds test this?
http://forums.dlink.com/index.php?topic=65889.0

[FrankTheTank]

QoS just simply doesn't work on the DIR-868L.  In my experience, with firmware 1.09 to 1.12, enabling QoS always results in a significant reduction in internet speeds.

[FurryNutz]

So does FW versions before v1.09 exhibit better results?

Can you post your speed results with QoS enabled and disabled like I did? Please.

QoS just simply doesn't work on the DIR-868L.  In my experience, with firmware 1.09 to 1.12, enabling QoS always results in a significant reduction in internet speeds.

[FrankTheTank]

The OP made it clear that 1.04 has the same problem.

I have a 55/5 connection.  Enabling QoS results in a 18-20% drop in connection speed.  DLINK was aware of this issue long before EOL, but decided not to work on a fix.