Changing MTU from 1428 to 1500 Yielding Speed Boost

[Lyondellic]

It's NYC lol. Every corner has some kind of pico cell and it's usually fed by preexistent fiber. NYC is loaded with dark fiber.

Still trying to figure out if I should just root and flash with that latest Gingerbread RUU or should I just keep it stock. I'm just trying to get the best possible data performance, don't really care too much about the features of the ROM.

Seems like any further gains would likely be transparent given your 60+ Mbps Download speed. The only thing this may help with is if you are Downloading massive files. Just my .02 cents though.

 

[milan03]

This is true for 3G/4G, but not when the phone is connected to Wi-Fi. A max MTU setting of 1492 is needed when connected to a Wi-Fi network that originates from DSL.

That's simply not true. If you are using DSL connection your router is what handles the MTU automatically. After that everything else is 802.11 protocol not DSL.

 

[justavillain]

I changed it and my Upload on 3g is a lot higher than before but my download is a hair lower. I'm using the newest radio. I might go back down to 802

But my wifi is WAY better getting an avg of about 5000 kbps down and

I'm getting around 9mbps down and about 3 mbps up. My laptop is getting a 19mbps down and about 3 up.

Edit: I reflashed the 802 radio pack and I'm getting better downloads and lower uploads on 3g. Ill try tomorrow on 4g

 

[milan03]

I changed it and my Upload on 3g is a lot higher than before but my download is a hair lower. I'm using the newest radio. I might go back down to 802

But my wifi is WAY better getting an avg of about 5000 kbps down and

I'm getting around 9mbps down and about 3 mbps up. My laptop is getting a 19mbps down and about 3 up.

Edit: I reflashed the 802 radio pack and I'm getting better downloads and lower uploads on 3g. Ill try tomorrow on 4g

I flashed the latest radio and speeds totally sucked. I'm not sure if Verizon is purposely releasing slower radios bit I'm going back to last froyo if that's the case. All day been vertigo between 5-20mbps in places I used to hit 60mbps...

Sent from my ThunderBolt using Tapatalk

 

[rladkins25]

Ok I went in root explorer and did all was said but Idk I can't find mtu am I missing something

Sent from my ADR6400L using xda premium

 

[mohkg]

Ya i couldnt find it ,but then i searched it again and found it look for this
system props for the data modules it should be under it
Sent from my ADR6400L using XDA App

 

[Grnlantern79]

Also dowload and use wordpad ++ its free and I love the format it breaks the info up into

Sent from my ADR6400L using XDA App

 

[Lyondellic]

That's simply not true. If you are using DSL connection your router is what handles the MTU automatically. After that everything else is 802.11 protocol not DSL.

I am always learning, and welcome any information that points me in the right direction.

Many routers have a setting in which you can alter its MTU size. You can change this, but you must make sure that the MTU setting on your router is the same as on your PC. It does not matter too much if the MTU on your router is higher than that on your PC, but the setting in your router must not be less than the MTU on your PC.

Router/Ethernet (Wired PC) MTU >> Max setting of 1500 (Cable) and 1492 (DSL), but the maximum setting may not be the optimal setting for all. Testing needs to be performed to determine the optimal MTU on a connection by connection basis. The easiest method to do this is via a desktop computer and the command line.

1. Open a DOS prompt

2. Type 'ping www.google.com -f -l 1492' and then press the Enter key (Note: Do not confuse 1492 with Cable/DSL Max MTU at this point)

3. Reduce the packet size in increments of 10 (e.g. 1482, 1472, 1462, 1452) until you find a packet size that consistently does not fragment

4. Begin increasing the packet size from this number in small increments (e.g. by +1 or +2) until you find the largest size that does not fragment

5. Add 28 to number of largest size that does not fragment to determine your optimal MTU setting (For example, if largest packet size from ping tests that does not fragment is 1472, add 28 to 1472 to get optimal MTU setting of 1500)
___________________________________________________________________

Wireless 802.11 networks have higher Mandatory MTU's than 802.3 Ethernet networks. The maximum frame size of 7935 bytes in 802.11n is now serialized in about 7.5 times less than a 1500 byte frame in 802.11b.

Table: 802.11 Evolution and Mandatory MTUs
Technology.....Rate.............MTU
802.11b.........11 Mb/s.........2272
802.11g.........54 Mb/s.........2272
802.11n.........432 Mb/s ......7935

So, what I take from the above is that the MTU setting employed between the wireless access point and client is much higher than 1492/1500 used for wired Ethernet connections to the router and router connections to the ISP. Is my understanding of this correct? This is new to me, and I welcome any additional information or examples.

 

[ghstrdr1985]

So after reading all this, what would be the ideal best setting for 4G/LTE networks?

 

[yareally]

You can test for how high you can set your MTU by doing the following:


Open the terminal and enter this following command:

ping -c 5 -s [trial MTU number] google.com (or some other url)

Start with a MTU of say 1500 and go up or down depending on the message (if it tells you that the packets are being fragmented, you need to go down--you want the highest setting that doesn't cause packet fragmentation). To figure out your MTU from the ping info, add 28 to the highest number that worked (packet size + 28).

Example: ping -c 5 -s 1500 google.com

If this does not cause fragmentation, try 1532 and keep going up by mutiples of 8.


Keep repeating until you get fragmented packets and then drop back down to the lowest MTU that did not.

 

[Trigster]

Newb here:

So what radio is in the OTA GB update?

I noticed pretty much solid cap of 10mbps upon applying the GB update a week ago. Prior to the update it was pulling 23-29mbps off the towers, and after I couldn't break 10mbps.

 

[milan03]

Newb here:

So what radio is in the OTA GB update?

I noticed pretty much solid cap of 10mbps upon applying the GB update a week ago. Prior to the update it was pulling 23-29mbps off the towers, and after I couldn't break 10mbps.

I'm also seeing 50-60% decrease after applying Gingerbread radios. I think it has something to do with buffer size on the kernel level.

 

[yareally]

Newb here:

So what radio is in the OTA GB update?

I noticed pretty much solid cap of 10mbps upon applying the GB update a week ago. Prior to the update it was pulling 23-29mbps off the towers, and after I couldn't break 10mbps.

Have you thought about asking in a posting that is more on topic? There's a billion "slow speed/throttling" posts already on this board and some in the last week (one specificially about GB)

 

[silvrhand]

I would look more at TCP window size issues, more than I would MTU, tcp window size has a much larger impact on the performance of higher latency devices. Adjusting your MTU by less than 10% isn't going to give a 100% increase in speed.

When it's too good to be true, it usually is.

 

[ghstrdr1985]

I would look more at TCP window size issues, more than I would MTU, tcp window size has a much larger impact on the performance of higher latency devices. Adjusting your MTU by less than 10% isn't going to give a 100% increase in speed.

When it's too good to be true, it usually is.

So can the tcp values be adjusted in the build.prop file as well? If not, how can they be adjusted?

Sent from my ADR6400L using xda premium

 

[milan03]

So can the tcp values be adjusted in the build.prop file as well? If not, how can they be adjusted?

Sent from my ADR6400L using xda premium

I'm pretty sure Rwin isn't in build.prop.

 

[swieder711]

Interesting thread.

Performance changes seem to be all over the place with both improvements and degradation reported on the latest 906 radios.

Increaing MTU to 1500 looks interesting. Hard to imagine > 10% performance gains.

I wonder how reliable the performance test really are.

just my .02

 

[sdnick484]

So can the tcp values be adjusted in the build.prop file as well? If not, how can they be adjusted?

Sent from my ADR6400L using xda premium

Since the phone is just running a version of Linux, you can echo new values to the correct places in the /proc file system. Install a terminal program, su to root and have at it. For details on what to change, check here.

E.g.
# cat /proc/sys/net/core/rmem_max
262144
# echo 524288 > /proc/sys/net/core/rmem_max
# cat /proc/sys/net/core/rmem_max
524288

By default TCP Window Scaling is enabled, so setting a (reasonable) higher buffer may very well help for high latency links. TCP will automatically climb to the highest size that's beneficial (in theory). You should generally set it to slightly above your Bandwidth Delay Product (BDP). This will be somewhat arbitrary since it's our response times will vary significantly depending on where you're connecting and tons of other factors.

Anyways, I'm curious to see how this goes. I just doubled mine (my default was 256K using the default kernel in Liquid Sense 2.6). Keep in mind these changes are lost when you reboot. There's ways to make them persistent, but this is probably better for testing.

 

[Andrew C]

You can test for how high you can set your MTU by doing the following:


Open the terminal and enter this following command:

ping -c 5 -s [trial MTU number] google.com (or some other url)

Start with a MTU of say 1500 and go up or down depending on the message (if it tells you that the packets are being fragmented, you need to go down--you want the highest setting that doesn't cause packet fragmentation). To figure out your MTU from the ping info, add 28 to the highest number that worked (packet size + 28).

Example: ping -c 5 -s 1500 google.com

If this does not cause fragmentation, try 1532 and keep going up by mutiples of 8.


Keep repeating until you get fragmented packets and then drop back down to the lowest MTU that did not.

I got all the way up to 10,000 MTU before fragmentation started to occur! I didn't change the build.prop, I was just curious to see what the command did.

 

[milan03]

Since the phone is just running a version of Linux, you can echo new values to the correct places in the /proc file system. Install a terminal program, su to root and have at it. For details on what to change, check here.

E.g.
# cat /proc/sys/net/core/rmem_max
262144
# echo 524288 > /proc/sys/net/core/rmem_max
# cat /proc/sys/net/core/rmem_max
524288

By default TCP Window Scaling is enabled, so setting a (reasonable) higher buffer may very well help for high latency links. TCP will automatically climb to the highest size that's beneficial (in theory). You should generally set it to slightly above your Bandwidth Delay Product (BDP). This will be somewhat arbitrary since it's our response times will vary significantly depending on where you're connecting and tons of other factors.

Anyways, I'm curious to see how this goes. I just doubled mine (my default was 256K using the default kernel in Liquid Sense 2.6). Keep in mind these changes are lost when you reboot. There's ways to make them persistent, but this is probably better for testing.

I've been trying 524288 for the past 5 days, didn't see any performance increase at all. Should I double that?!