[ANSS-netops] rtpd max clients [Memphis]

Mon May 17 18:09:56 UTC 2010

Mitch,

Like others have said, rtpd has worked well for us in Menlo.  We have 
over 100 stations being served by two old Dell servers (one is still 
a Pentium 3!).  The only reason we have two servers is because we 
have so many stations.  We parse the stations according to geographic 
area and telemetry methods so that if one machine fails, we're still 
covered with the other.

If Ian at Reftek didn't answer all your questions or if you have 
other questions related to our set up, contact Lynn Dietz who is our 
rtpd master (dietz at usgs.gov).

Dave

At 01:54 PM 5/13/2010, Jamison Steidl wrote:
>We had 30 stations at our Garner Valley site, all coming into a SUN Netra 1U
>server via a single omni antenna and WiLAN radio, with no problems. My
>recollection is that I was surprised at how small the CPU load was, and the
>fact that 3 channels x 30 stations x 200 sps in real-time was no problem for
>the radio link. RTCC worked great for command and control of all the
>stations from back at UCSB.
>
>Cheers,
>Jamie
>
>
> > From: <mwithers at memphis.edu>
> > Reply-To: <mwithers at memphis.edu>
> > Date: Thu, 13 May 2010 07:04:27 -0500 (CDT)
> > To: Ian Billings <I.Billings at reftek.com>
> > Cc: "anss-netops at geohazards.usgs.gov" <anss-netops at geohazards.usgs.gov>
> > Subject: Re: [ANSS-netops] rtpd max clients [Memphis]
> >
> >
> > Thanks to all those who responded and to those who are still thinking about
> > it.  Just to be clear, we're not having problems.  Rather I'm trying to
> > learn from the experience of others as we plan for the future.  What I've
> > learned so far is very useful.
> >
> > Mitch
> >
> > Center for Earthquake Research and Information (CERI)
> > University of Memphis                Ph: 901-678-4940
> > Memphis, TN 38152                   Fax: 901-678-4734
> >
> >
> > On Wed, 12 May 2010, Ian Billings wrote:
> >
> >> Mitch,
> >>
> >> We were waiting to reply to this email in case others may have had some
> >> input so we could learn some more about how users not only use our
> >> software but how users would like to see our software perform.
> >>
> >> It is recommended that the latest RTPD version and data clients are
> >> used.  REF TEK sends a notification to users registered on our
> >> http://support.reftek.com site when new versions are available, or
> >> alternatively one can check the site, on regular basis, for updates.
> >>
> >> The rtpd.ini file entry for Max Clients relates to data clients like
> >> RT_Display, RTPC etc.  It does not relate to RTP (130 or 72A)
> >> connections or the Archive connection.  RTP connections have priority
> >> over Client connections therefore always serviced first.  The Archive
> >> thread, if present, is also of similar priority to RTP connections.  The
> >> RTP/RTPD connection is full error correcting to allow 100% data.  If a
> >> link between a RTP/RTPD connection is slow or intermittent then RTPD
> >> will create more overhead for the system as it tries to keep the data
> >> flow 100%.
> >>
> >> RTPD can handle as many RTP connections as the platform performance will
> >> allow, easily in the hundreds of connections if the links are stable.
> >> The observed slow down in performance may come from a Client connection
> >> like RT_Display as these clients use other resources in the system to
> >> display the data etc.  Intermittent or slow links increase overhead for
> >> RTPD as mentioned but rarely would there be an observed slowdown in
> >> performance, instead delayed or gapped data flow from the intermittent
> >> station.
> >>
> >> RTPD to RTPD connections are also of small overhead, rtp2rtpd for
> >> example. However the clients connected to these additional RTPD's may be
> >> the area where you would see performance problems if the clients are
> >> using other system resources heavily.  REF TEK would not recommend
> >> multiple instances of RTPD on the same machine unless different ports
> >> were used via rtp2rtpd.   UDP 2543 is the RTP Fixed data port and this
> >> should have a single RTPD to connect to, the primary.
> >>
> >> The only other time you may see problems in performance is when the
> >> archive has reached its purge threshold and starts using the buffer
> >> space to keep current data while deleting older data.  There is a
> >> section in the rtpd.ini file to help stream line this purging process,
> >> rather than deleting single files, blocks of files or an allotted
> >> archive space can be deleted in one go.
> >>
> >> I hope this helped and please let us know if you have any further
> >> questions.
> >>
> >> Ian billings.
> >> Customer Service Engineer
> >> Refraction Technology Inc
> >> 1600 10th Street Plano TX 75074 USA
> >> Ph 214 440-1265 or 214 440-1289 (direct)
> >> http://support.reftek.com
> >>
> >> -----Original Message-----
> >> From: anss-netops-bounces at geohazards.usgs.gov
> >> [mailto:anss-netops-bounces at geohazards.usgs.gov] On Behalf Of
> >> mwithers at memphis.edu
> >> Sent: Tuesday, May 11, 2010 7:40 AM
> >> To: anss-netops at geohazards.usgs.gov
> >> Subject: [ANSS-netops] rtpd max clients
> >>
> >>
> >> Anyone running a monster RTPD with many clients?  I'm running rtpd
> >> 2.1.8.0
> >> on a SunFire V210 with Soalaris 10 and have MacClient configured to 16.
> >> Actual connections are 7 DAS's, one instance of rtp2rtpd, rtcc, and rtpm
> >> (I assume a client is anything that connects to the port whether its
> >> a das or software).
> >>
> >> In people's experience how many clients can you have before it starts
> >> bogging
> >> down if we assume 100BaseT internet?
> >>
> >> Is it better to run mulitple instances of rtpd on the same machine with
> >> different ports and relatively small MaxClient or is it better to run
> >> one big one with a large MaxClient?
> >>
> >> Does running multiple instances of rtpd on a single host cause problems?
> >>
> >> Can mulitple instances of rtpd be aggregated into a single rtpd via
> >> rtp2rtpd
> >> that doesn't have the overhead of a bunch of DAS connections?
> >>
> >> Ok, some of these questions may be more appropriate for reftek support
> >> but its still useful knowing how people have things lashed up.
> >>
> >> Mitch
> >>
> >> Center for Earthquake Research and Information (CERI)
> >> University of Memphis                Ph: 901-678-4940
> >> Memphis, TN 38152                   Fax: 901-678-4734
> >>
> >> _______________________________________________
> >> ANSS-netops mailing list
> >> ANSS-netops at geohazards.usgs.gov
> >> https://geohazards.usgs.gov/mailman/listinfo/anss-netops
> >>
> >>
> >>
> > _______________________________________________
> > ANSS-netops mailing list
> > ANSS-netops at geohazards.usgs.gov
> > https://geohazards.usgs.gov/mailman/listinfo/anss-netops
>
>
>_______________________________________________
>ANSS-netops mailing list
>ANSS-netops at geohazards.usgs.gov
>https://geohazards.usgs.gov/mailman/listinfo/anss-netops