[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[LDM #VPY-764086]: Performance issues > 6.6.0

This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.

  • Subject: [LDM #VPY-764086]: Performance issues > 6.6.0
  • Date: Tue, 10 Apr 2007 15:07:37 -0600
Daryl & Gilbert,

We've managed to convince ourselves that the apparent
increase in the CPU load average due to upgrading to
LDM 6.6 from LDM 6.4 is just that: an apparent increase
that doesn't truly describe the situation.  Upstream
LDM 6.4 processes flushed the connections to downstream
LDM-s unnecessarily; consequently, they were often in
a state where they were waiting on a socket and not
ready to run.  Upstream LDM 6.6 processes, however,
flush their connection to downstream LDM-s only when
necessary; consequently, they are more often ready
to run.  If you could see the state of upstream LDM
processes, then you would see that relatively more of
the LDM 6.4 processes are in the "waiting" state than
LDM 6.6 processes and, conversely, more LDM 6.6
processes are in the "ready to run" state than than
LDM 6.4 processes.  The CPU, on the other hand,
executes only one upstream LDM process at a time and
does so as much as necessary to transmit the required
amount of data in the required amount of time in both
cases.  Therefore, the difference between LDM 6.4 and
LDM 6.6 isn't that LDM 6.6 puts more load on the CPU
than LDM 6.4, but that the upstream LDM 6.6 processes
have a different state distribution than the upstream
LDM 6.4 processes.

Regards,
Steve Emmerson

Ticket Details
===================
Ticket ID: VPY-764086
Department: Support LDM
Priority: Normal
Status: Closed