Skip to Content.
Sympa Menu

perfsonar-user - [perfsonar-user] TCP retransmissions

Subject: perfSONAR User Q&A and Other Discussion

List archive

[perfsonar-user] TCP retransmissions


Chronological Thread 
  • From: Tao Zhang <>
  • To: "" <>
  • Subject: [perfsonar-user] TCP retransmissions
  • Date: Tue, 18 Jul 2017 19:21:49 +0000
  • Accept-language: en-US
  • Ironport-phdr: 9a23:fJBoUheuu6eutWp8+Van6ik9lGMj4u6mDksu8pMizoh2WeGdxc2zYx7h7PlgxGXEQZ/co6odzbGJ4+a9ASQp2tWojjMrSNR0TRgLiMEbzUQLIfWuLgnFFsPsdDEwB89YVVVorDmROElRH9viNRWJ+iXhpW1aJhKqfxJ4POruHYjblYGqzO2o05zVfwhSgjehO/V/IAj85VHJu9MYmoxkI7x01wDEuFNJffhb32VlOQjVkhrhsJSe5plmphhZvLoT98dGXu2uQ60/C45YBTIpG2s84sTxtBSFRgLZtShUaXkfjhcdW1uN1xr9RJqk6iY=

 

Hi,

 

We have newly installed the latest version of perfSONAR on a Dell R430 (new box).

 

Specs:

Processor #1: 3.00GHz (8 cores)
Memory: 31.90GB
NIC #1 Speed: 1Gbit/s
NIC #1 MTU: 9000
NIC #2 Speed: 10Gbit/s
NIC #2 MTU: 9000

 

Operating System: CentOS Linux 7.3.1611 (Core)
Kernel: Linux 3.10.0-514.26.2.el7.x86_64

 

 

 

We are conducting the throughput test between the new box and the rest of our perfSONAR boxes, which are running CentOS 6.9 (old box).

 

Processor #1: 2.39GHz (8 cores)
Memory: 11.89GB
NIC #1 Speed: 10Gbit/s
NIC #1 MTU: 9000

 

Operating System: CentOS 6.9 (Final)
Kernel: Linux 2.6.32-696.3.1.el6.x86_64

 

 

We have set up the host parameters as following:

 

net.core.wmem_max = 268435456

net.ipv4.tcp_rmem = 4096 87380 134217728

net.ipv4.tcp_wmem = 4096 87380 134217728

net.core.netdev_max_backlog = 250000

net.ipv4.tcp_no_metrics_save = 1

net.ipv4.tcp_congestion_control = htcp

net.ipv4.conf.all.arp_ignore=1

net.ipv4.conf.all.arp_announce=2

net.ipv4.conf.default.arp_filter=1

net.ipv4.conf.all.arp_filter=1

 

Our throughput can reach 10G, which is good.

 

The problem is that we are having the TCP retransmission issues between the new box and old box.

 

 

The tests between old boxes have no issue, only the test between new box and old box are having TCP retransmission.

 

Is that related to the different setting of queue discipline?  In Centos 7, the default qdisc is fq and for Centos 6, it is pfifo_fast.

 

Anyone has the similar issue?

 

Thanks

 

 

Tao Zhang

 Network Engineer | CANARIE Inc | 45 O'Connor St., Suite 500, Ottawa, ON K1P 1A4 │CANARIE NOC: 613-944-5612│www.canarie.ca

 

 

 

 

 




Archive powered by MHonArc 2.6.19.

Top of Page