[mythtv-users] Multiplexing vs multiple cards
oliver.greg at gmail.com
Thu Nov 4 18:05:45 UTC 2010
On Thu, Nov 4, 2010 at 1:00 PM, Douglas Peale <Douglas_Peale at comcast.net> wrote:
> On 11/04/2010 10:41 AM, Michael T. Dean wrote:
>> On 11/04/2010 01:16 PM, Raymond Wagner wrote:
>>> On 11/4/2010 13:11, Douglas Peale wrote:
>>>> BTW I was unable to find the option for real-time comm-flag in the menus. The only thing I found was "Enable experimental
>>>> speedup of commercial detection". Is that it?
>>> No. That intends to speed up commflagging by running the analysis on a half scale (quarter pixels) video.
>> Start Auto-Commercial Flagging jobs when the recording starts
>> If set and Auto Commercial Flagging is ON for a recording, the flagging job will be started as soon as the recording starts.
>> NOT recommended on underpowered systems.
>> (or in trunk:
>> Start auto-commercial-detection jobs when the recording starts
>> If enabled, and Auto Commercial Detection is ON for a recording, the flagging job will be started as soon as the recording
>> starts. NOT recommended on underpowered systems.
>> Note, though, that the option becomes mostly meaningless when you factor in back-to-back recordings.
>> Instead, setting the "Maximum simultaneous jobs on this backend" and "CPU usage" and "Job Queue start time" and "Job Queue end
>> time" appropriately will be useful for people who are attempting to run underpowered backend systems.
> Thank you.
> In my case I am not running an under powered system (i7 with 6 GB of memory).
> Thinking about this, the read may be avoided by the data still being in the cache when the comment flagger goes to read it. I
> think I'll give that a try. It should improve a disk limited system as long as the flagger does not get too far behind.
My i7 records and commflags 6 streams (usually only 4 with commercials
though), and the cpu stays mostly idle. I am not sure how the
commflagging does it's thing, but I do see spikes at times for brief
moments, but overall as long as they run with the recorders, the
system is very much in check. I moved my concurrent jobs to 6 just to
make sure that never happens though because it kills my virtual
machines almost to the point of unusability when the array thrashes.
More information about the mythtv-users