[Help] Newer Switch = Slower Speeds

B

bekax5

Limp Gawd
Joined
Jun 4, 2012
Messages
132
Hello everyone.

I have made a switch upgrade since I had one with a few years old (7), its web config was old and non-working with nowadays browsers, and I was expecting to get better performance overall with a fresh new device.
After watching a lot of reviews and stuff, I found a nice deal on a D-Link 1510-28 to change my old Linksys SRW-2024.
So on a first view I enabled Jumbo Frames as I had in the Linksys, and set some IPs.


Linksys SRW-2024 VS D-Link 1510-28:

Tested on iperf:
Linksys: 0-5s = 920Mbit/s | 5-10s = 922Mbit/s
D-Link: 0-5s = 490 Mbit/s | 5-10s = 860Mbit/s

Tested on windows File Transfer (4GB file):
Linksys: ~110MB/s
D-Link: ~68MB/s

Do you think I have something misconfigured or shouldn't I be expecting more from a 400€ switch?
 
Disable jumbo frames unless you have a very good reason to use it.
Are you running the latest firmware?
//Danne
 
diizzy said:
Disable jumbo frames unless you have a very good reason to use it.
Are you running the latest firmware?
//Danne
Click to expand...

Yes, latest updates.

Well, I thought lowering CPU usage and possibly better throughputs would be a good reason ?
In any case I get about the same speeds without any Jumbo Frames.
 
Are you running this hard wired or over wireless. What type of wiring are you running. Cat5, Cat5e, or Cat6.
 
Both Cat5E and Cat6.
All wired at 1Gbit.

I wouldn't guess cables are the problem since the same cables are being used on both switches.
 
just curious why buy D-link when spending 400 Euro's....

Why not get an HP or Dell or something...
 
Have you updated to the latest NIC drivers (on both ends)?
150 Euros is a pretty good deal, Zyxel GS1910-24 is about 110 EUR in .de which is a very good deal.

Quick test Client (Win7) --> WDR3600 (switch only) --> Zyxel GS1910-24 --> Server (FreeBSD 11-CURRENT)
938 Mbits/sec both ways...
//Danne
 
Last edited:
Updated already.
I have Intel Gigabit CT on both ends.

For the price I was thinking about selling my old Linksys SRW2024 allowing me to get this D-Link almost for "free", but after seeing this performance, I'm thinking twice!
 
Definitely something funny going on there. Check around in the GUI as some of the newer gear have "green" modes which might be holding it back from doing full speed. I have a 10 year old dlink 24 port that can do the full 100MBps+ between two ports. Make sure it's running full duplex as well.
 
bman212121 said:
Make sure it's running full duplex as well.
Click to expand...

What he means to say is make sure both sides, the switch ports and the PCs, are set to auto/auto. Do not ever hard code any speeds over 100Mb and seriously question why you're doing it then.
 
On paper that dlink wins, but unless there is some stray setting limiting performance then keep your Linksys. Paper specs vs reality.
 
I mean, both are switches...
At least close to default they should both give near line rate, or at least I would expect both to give 900+Mbits/s and 90+MB/s on samba...

I'll leave some pics at the end of the week because I'm not at home now.
 
bekax5 said:
I mean, both are switches...
Click to expand...

Yes, but not all switches are created equal. Lots of things can affect switch performance, shared buffer size, cpu speed, #of asics, shared fabric etc. Check your end points, check the switch for strange settings like "green" settings etc. Honestly it just sounds like it cannot perform as fast as the Linksys.
 
I am really sad with this new 400€ switch not being able to perform as good as a 10 years old linksys costing 150€
But I still hope to figure out any strange configuration that I haven't notices. And let's just say that there's a lot of configs :p
 
Would be interesting to know what hardware it uses, my Zyxel GS1910 (which actually is cheaper) uses a Vitesse platform.
//Danne
 
bekax5 said:
I am really sad with this new 400€ switch not being able to perform as good as a 10 years old linksys costing 150€
But I still hope to figure out any strange configuration that I haven't notices. And let's just say that there's a lot of configs :p
Click to expand...

Clear the config to OOB and run it stateless to see how it performs.

Switch performance has nothing to do with how new it is, but what is inside hardware wise.
 
schizrade said:
Clear the config to OOB and run it stateless to see how it performs.

Switch performance has nothing to do with how new it is, but what is inside hardware wise.
Click to expand...

Well honestly haven't changed much!

I did change IP/mask, enabled SNMP which is at the moment being used to monitorize, once each minute, about 10 ports' total traffic, and mirror one of the ports that won't pass more than 30Mbit/s of traffic to another.
Oh, and tried both with and without jumbo frames.

This should easily be taken care by any basic managed switch. (as it was by the previously mentioned 7+years old SRW2024)

But yea, as soon as I am in personal contact with the switch, I will reset to defaults.
 
bekax5 said:
Well honestly haven't changed much!

I did change IP/mask, enabled SNMP which is at the moment being used to monitorize, once each minute, about 10 ports' total traffic, and mirror one of the ports that won't pass more than 30Mbit/s of traffic to another.
Oh, and tried both with and without jumbo frames.

This should easily be taken care by any basic managed switch. (as it was by the previously mentioned 7+years old SRW2024)

But yea, as soon as I am in personal contact with the switch, I will reset to defaults.
Click to expand...

Yeah it sounds like something is config wrong, or it is just not a very good switch. My $250 HP V1910-16G outperforms that dlink.
 
i had a cisco branded switch that barely did 100MB/s speeds steady on gigabyte, just a rebranded piece of crap it seemed.
 
Since you found what seems to be a pretty hot deal on that, are you sure the switch is genuine and not a fake?
 
It's EoL, from what I can tell it's been around ~290 EUR for the last year or so.
//Danne
 
munkle said:
Since you found what seems to be a pretty hot deal on that, are you sure the switch is genuine and not a fake?
Click to expand...

It is a genuine. It did bring all the original manuals, and firmware through console port and web management seems as original dlink.
Also, I did open the switch and it corresponds to what I was expecting from online images.

diizzy said:
It's EoL, from what I can tell it's been around ~290 EUR for the last year or so.
//Danne
Click to expand...

EOL or not, it is (should be) at the moment one of the top smart managed stackable switches from dlink. Should not be that bad XD

I'm waiting for the weekend to test it further...
 
Came back home and I found a new firmware, updated and I did reset to defaults.
But still, the issue persists.

I took some prints, could this be the switch dropping packets?

D-Link only.


Top CMD is Linksys. Bottom CMD is D-Link.


As you can notice, the D-Link starts to really send packets at "full speed" after 2,5 seconds!
Strange behaviour ?
 
Uhm... Give iperf3 a go instead. I'm just curious to see if you get the same results.
//Danne
 
I guess I have to take some time and compile it for windows.

In any case, I tried with different Network cards, and I got better results.
Transferring files through samba on my Macbookpro, I got speeds between 920-970Mbit/s

So I'm starting to guess there's some misconfig between the two Intel cards.
I see that they are actually pushing ~970Mbit/s (in windows resource monitor) but iperf misses about 100Mbit/s that I imagine are taken as lost packets or something.

Also with the Macbookpro I didn't get the slowdown at the begining of the transfer.
 
You must log in or register to reply here.
Back
Top