Degraded network application performance over loopback interface in RHEL 6.6 and later

Solution Verified - Updated -

Issue

  • My team is in the middle of upgrading our RHEL systems from6.4 to 6.6 when we came across this odd behaviour.
  • We have an old in-house developed file transfer utility, written in C, that has shown a sudden decrease in transfer speeds when the source and destination host is the same machine, when the host is patched to 6.6.
  • On 6.4, this program can transfer an 18GB file in about 3-4 minutes. The same program, on the same machine patched to 6.6, now takes over 5 hours to transfer the same size file.
  • Transfer speeds between patched hosts, unpatched hosts, and combinations thereof are NOT affected. The transfer speed hit ONLY occurs when the application tries to send and receive from the SAME host.
  • Testing with scp to copy a file in a similar manner does NOT show a speed hit. This would seem to point to an issue with our application, but the circumstances seem to point to some interaction between the application and the patched system, since the application itself has not been changed, either.

Environment

  • Red Hat Enterprise Linux 6.6 and later (kernel-2.6.32.504.el6 and later)
  • TCP/IP networking over loopback interface lo
    • Note: Sending and receiving to an IP address of a NIC on the same host will result in the kernel using the loopback driver.

Subscriber exclusive content

A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.

Current Customers and Partners

Log in for full access

Log In