Tutor
•
7 Messages
devices connected to each other not getting full 1Gbps with Pace 5268ac gateway with iperf tests
my internet has been fraught with issues and thought I'd address them one at a time. The one issue I'd like to bring up here is that regarding the slow connection between two clients in a LAN when mediated by my Pace 5268AC. When these same devices are connected to each other through a switch or router that are 1Gbps capable, I get around 800-1000Mbps, whereas when connecting with the gateway alone, I get anywhere from 100-500Mbps, confirmed by a series of iperf tests that were run as shown below.
To specify:
Example 1(desktop1 and desktop2 connected to a router):
Client connecting to 192.168.2.108, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 56916 connected with 192.168.2.108 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 1001 MBytes 839 Mbits/sec [root@admin ~]$ iperf -c 192.168.2.108 ------------------------------------------------------------ Client connecting to 192.168.2.108, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 56918 connected with 192.168.2.108 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 1.06 GBytes 906 Mbits/sec [root@admin ~]$ iperf -c 192.168.2.108 ------------------------------------------------------------ Client connecting to 192.168.2.108, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 57048 connected with 192.168.2.108 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 1.04 GBytes 896 Mbits/sec
Example 2(same desktop 1 and desktop2 connected to Pace 5268ac gateway):
(n.b. different subnet using gateway's dhcp)
[root@admin ~]$ iperf -c 172.16.0.12 ------------------------------------------------------------ Client connecting to 172.16.0.12, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 52874 connected with 172.16.0.12 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 148 MBytes 124 Mbits/sec [root@admin ~]$ iperf -c 172.16.0.12 ------------------------------------------------------------ Client connecting to 172.16.0.12, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 52876 connected with 172.16.0.12 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.0 sec 146 MBytes 122 Mbits/sec [root@admin ~]$ iperf -c 172.16.0.12 ------------------------------------------------------------ Client connecting to 172.16.0.12, TCP port 5001 TCP window size: 85.0 KByte (default) ------------------------------------------------------------ [ 3] local 192.168.2.47 port 52878 connected with 172.16.0.12 port 5001 [ ID] Interval Transfer Bandwidth [ 3] 0.0-10.1 sec 146 MBytes 122 Mbits/sec
-----
I had tried to troubleshoot this with multiple techs, but upon reading this forum, I'm more confident in finding an answer here than on the phone and in-person. The tests were performed with the same cables and systems with the only major different variable being the device mediating communication between the two devices, so that the comparisons are as apples to apples as they can get.
I had also ran these tests using multiple devices in a similar fashion where two devices with the same cables will have their results compared between being connected to a third party router and the Pace 5268ac (eg Desktop 3 and Laptop 1 connected to router VS Desktop 3 and Laptop 1 connected to Pace 5268ac). I have also tried using different port combinations between the four to identify defective ports, with the assumption that at least two of the ports on the Pace 5268ac are not defective. And lastly, I had ran these tests with both of the two client-server options available (eg Desktop 1 as server vs Desktop 1 as client). In so doing, I found that these variations had no material effect on the original findings aside from confirming them.
Uverse, as of today, is shipping a replacement gateway, but my concern is that this will not resolve the issue and would appreciate any advice regarding configuration or the parameters of the tests to conduct to measure the health of the device and its capacity to perform the simple task of moving information between at least two 1Gbps capable devices at the 800-1000Mbps bandwidth I get elsewhere. Thank you.
mibrnsurg
Expert
•
20.4K Messages
8 years ago
@comptroller_jim Uverse RGs, combination modem/router, designed for Uverse triple play, Uvoice, IPTV and Internet (not necessarily wi-fi). This must be used for ATT authorization and connection to ATT servers.
Many users add a 3rd party 'n or ac' router behind the RG, in DMZ zone on 5268, for better wi-fi and full router features. 😉
Chris
__________________________________________________________
Please NO SD stretch-o-vision or 480 SD HD Channels
Need Help? PM ATT Uverse Care (all service problems)
ATT Customer Care(billing and all other problems)
Your Results May Vary, In My Humble Opinion
I Call It Like I See It, Simply a U-verse user, nothing more
0
0
comptroller_jim
Tutor
•
7 Messages
8 years ago
Right. My concern is that, even though using a third party router can get around the limitations of the 5268ac or any other RG for communication between clients within the LAN, said third party router eventually needs to be connected to the RG via DMZ plus, and that usually involves the third party router being plugged into one of the four non-performant ethernet ports on the gateway(ie. 3rd party router's WAN port ======> RG's LAN port [1,2,3,4]). As such, it could mean that the wild fluctuations and declines in bandwidth that I've experienced over the internet could be in part caused by the construction of the gateway itself.
But I'd be grateful for anyone who can lead me in the right direction to bypass this issue completely or for showing how tenuous the link is between my tests and my conclusions.
0
0
JefferMC
ACE - Expert
•
35.1K Messages
8 years ago
Reading through your logs of your tests, it seems to me that you're testing apples and oranges.
The switch test was a layer-2 test fo the switch.
It appears to me that the test using the 5268ac was a layer-3 test, which will not perform as well. I say this because of this line in your log:
This seems to indicate that it's using the outbound interface 192.168.2.147 to communicate to 172.16.0.12, which will require routing (unless you've done something "interesting" with the netmasks). To be fair, it should use the local 172.16.0.xx IP address because routing work to the process that switching does not.
0
0
mibrnsurg
Expert
•
20.4K Messages
8 years ago
@comptroller_jim What speed do you get in a gigabit speed test online w/a computer connected to an ethernet port on the 5258?
Chris
__________________________________________________________
Please NO SD stretch-o-vision or 480 SD HD Channels
Need Help? PM ATT Uverse Care (all service problems)
ATT Customer Care(billing and all other problems)
Your Results May Vary, In My Humble Opinion
I Call It Like I See It, Simply a U-verse user, nothing more
0
0
comptroller_jim
Tutor
•
7 Messages
8 years ago
I totally see what you mean by that. I should have narrowed in in thread on the tests that to me seem more apples to apples -- namely, testing the two clients(as iperf server and iperf client) connected to the 5268AC's LAN ports vs the two same two clients connected to the router and/or switch.
test using Pace gateway LAN ports(1,2)(results sim. for using other two-port combo)
test using router's LAN ports(sim. to above, ports 1,2 and combos of all ports)
In the above example, they're the same two clients used in that particular series of tests--as noted by the fact that although they have different ips in the two tests, the two clients are in the same subnet per test.
So yeah, ultimately, I'm trying to find the bottlenecks in my WAN and bugs to clear and squash them as much as I can, but given my findings, it seems very easy to conclude that the Pace 5268AC is not delivering on the 1Gbps promise. If i got at least the 800ish Mbps I got/get with the router tests(btw, I'm submitting the least flattering test for the router as other tests have yielded up to 1.05Gbps...dunno why), then I wouldn't flinch. But with the discrepancy between the 850Mbps of the router and the 160Mbps of the gateway, (I didn't include more tests with the switch involved, which also delivered the rated 1Gbps with a margin of 150Mbps) I just had to say something to get informed opinions like yours to either let me know that I'm way off base or that maybe I found something that needs to be called for examination by AT&T.
I found a workaround for the LAN by having my router do all the routing for me. But the part I hate the most is that I still need to make use of one of the LAN ports of the gateway to access the internet, as even the gateway's namesake declared. Thanks for your response and hopefully I can eventually either understand what accounts for this discrepancy and/or how someone in Uverse needs to make some calls.
By the way, the test above was with the replacement Pace 5268AC I received just yesterday, so my gateway doesn't at least so far seem to be defective, but the issue persists across units.
0
0
JefferMC
ACE - Expert
•
35.1K Messages
8 years ago
Assuming your netmasks are 255.255.0.0 for the 172.16.0.0 address pool (they are, right?), your test appears valid, and I'm surprised at the result. I don't have a 5268ac to replicate your test, unfortunately.
0
0
comptroller_jim
Tutor
•
7 Messages
8 years ago
0
0
JefferMC
ACE - Expert
•
35.1K Messages
8 years ago
AT&T calls a lot of shots in regards to what the firmware does (as opposed to what it would do if ARRIS were left to assign features). Evidence for that is that the 2WIRE/PACE firmware behaved much the same as the ARRIS firmware prior to ARRIS' purchase of PACE (yes the screens were different and the features were named different things, but otherwise...)
AT&T has reserved the 10.0.0.0/8 subnet for its own use. Stupid, yes, but it is reality.
I'd send a Private Message to @ATTU-verseCare to see if they think the 5268ac should be performing like that between LAN ports.
OTOH, switches are quite cheap. I recently bought a 5-port VLAN-capable Netgear Gigabit switch for $15 from NewEgg (after reward card, of course).
0
0