SUMMARY: Realtime OS on SUN?

From: J. Theuerkauf (joerg@wp-470-1.fl.bs.dlr.de)
Date: Thu Jul 30 1992 - 11:03:23 CDT


Hello sun-managers,

Thanks to all who answered to my request about realtime OSs on SPARCs.

My original message:

> Hi sun-managers,
>
> one of my colleagues looks for a VMEbus machine with realtime capabilities.
> Can anyone tell us about experiences with realtime operating systems (or extensions)
> on SPARC VME machines?
>
> Will SysV.4 (Solaris 2.0) serve realtime demands?
>
> The application deals with high volume data input on a couple of parallel ports.
>
> Thanks for any suggestion ...

I got answers from

hs@demeter.ukps.gwdg.de (Howard Schultens)
dunstan@gec-epl.co.uk (Dunstan_Vavasour)
MATTERN@ARIEL.LERC.NASA.GOV
field@cs.pitt.edu (Brian Field)
kalli!kevin@fourx.Aus.Sun.COM (Kevin Sheehan {Consulting Poster Child})
mcgraw@sunspot.sunspot.noao.edu (Robert McGraw)
stern@sunne.East.Sun.COM (Hal Stern - NE Area Systems Engineer)
davis@udecc.engr.udayton.edu (Thomas J Davis)
billr@saab.cna.tek.com (Bill Randle)
era@niwot.scd.ucar.EDU (Ed Arnold)

Some excerpts from their original replies:

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>Will SysV.4 (Solaris 2.0) serve realtime demands?

Yes. Talk to your local Sun Salesdroid. Mine slipped me a copy of an
internal Sun memo about the realtime capabilities of SunOS 5.

-- 
 Dunstan  | Systems Design Division           |dv@gec-epl.co.uk
 Vavasour | CEGELEC Projects                  |From JANET:
          | Rugby, Warwickshire, CV21 1BU, UK |dv%uk.co.gec-epl@uk.ac.uknet
          | Tel: (+44) 788 563535             |Fax: (+44) 788 560767

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Yes, Sun has a series of "white" papers on Solaris and real time. Ask your Sun distributor. The ones that I have are: Sun OS Multithread Architecture 1) Threads and Treads Programming for Applications 2) Sun OS Multithread Architecture 3) Writing Multithreaded Code in Solaris 2.0 Solaris 2.0 Multithreaded Kernel 4) Beyound Multiprocessing...Multithreading the SUn OS Kernel 5) Realtime Scheduling in SUN OS 5.0 6) Symmetric Multiprocessing in Solaris 2.0

I didn't find any of these papers useful (I'm a controls engineer, not a professional programmer). > >The application deals with high volume data input on a couple of parallel ports. There are many hardware/software vendors with products for the new Sun. Heurikon Corp. Madison, WI (608)831-0900 makes a MVEbus interface card for realtime audio appications.

We have also been talking to Lynx-OS. They claim that their next release will be POSIX complient and be able to run on the SUN. They have some kind of agreement/arrangement with SUN. > Duane Mattern (mattern@mars.lerc.nasa.gov) (216)433-8186 Sverdrup @ NASA Lewis, 21000 Brookpark Rd., Cleveland, OH 44135

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

VxWorks is quite a nice little package - check with Wind River Systems. > > Will SysV.4 (Solaris 2.0) serve realtime demands?

It has a realtime scheduling class, which will get you all of the CPU. There is some more work to be done if you want to use kernel resources as well. > > The application deals with high volume data input on a couple of parallel ports.

This is not necessarly a real-time application. data acquisition != real-time, as you can write a driver that will quite comfortably take in high data rates with no lossage if you a) have enough buffering or b) know how to write that kind of driver correctly.

Kevin Sheehan Unix Professional Services kalli!kevin@fourx.aus.sun.com

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Look at WindRivers VX works. Probably the best out their. phone(415) 748-4100 fax (415) 814-2010

Robert

-- ------------------------------------------------------------------------ Robert P. McGraw, Jr. System Manager INTERNET:rmcgraw@sunspot.sunspot.noao.edu National Solar Observatory/SP SPAN: 5355::RMCGRAW or NOAO::RMCGRAW Sunspot, NM 88349 USA PHONE: (505) 434-7038

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

from stern@sunne.East.Sun.COM (Hal Stern - NE Area Systems Engineer)

(a) a vme sparcengine board will run the vrtx real-time os, if you really need it. (b) solaris 2.0 has fixed priority scheduling, kernel pre-emption, and an execpted, bounded scheduling/dispatch latency of about 5 ms. (c) depending upon how the data must be written/consumed, you may not need a real time system -- just a good device driver. if the data must be captured and written to disk, you can use a smart driver that maps everything onto a raw disk partition -- then consume the raw partition later, from a user level program. if you need to actually modify the data as it comes it, and then write it to disk, then you might still be ok if you only have 1 process doing the collection. if you have several processes that must be scheduled according to inputs, you'll need a RT OS.

--hal -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Sun and System V.4 do have realtime extensions and may fit your environment, but the best realtime VME system you can buy, IMHO, is the Harris NightHawk.

NightHawks are built as realtime machines with full System V.4 symetric multiprocessing capabilities. There are NightHawks for just about every budget and application. The low end models have Motorola 68030 but everyother model uses either an 88100 or 88110, the latter rated at 70 SPECmarks per processor. Each processor can have up to 256KB cache and 64MB of local memory. Each machine has up to 256MB of global reflective memory. The newest machine, the 5800, uses a VME-64 and is rated at 55MB/sec.

The address for Harris in Germany is:

Harris Gmbh Computer Systems Division Schatzbogen 86 D-8000 Munchen 82 Ph. (49) 98.420.1166

One of the salesmen there is Reinhardt Wendt.

================================================================================ Tom Davis davis@udecc.engr.udayton.edu AT+T: 513-229-3171 Mail: Univ. of Dayton "The wise are pleased when they 300 College Park discover truth, fools when they Engineering Computer Center discover falsehood" Dayton, OH 45469-0229 -Gabirol's Observation

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ from billr@saab.cna.tek.com (Bill Randle)

Wind River Systems offers a port of their VxWorks real-time operating system for SPARC/VME machines. I'm sure they have some European reps, but I don't have that info. Their main office is in California: (415)748-4100, and Fax: (415)814-2010.

-Bill -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

The Solaris 2.0 Transition Q&A manual (910-2709-01) mentions these issues. In general, 2.0 cannot provide better than millisecond dispatch latency, but does provide several features such as an optional, fixed-priority process scheduler; a fully pre-emptible kernel and pre-emptive scheduling; and high-res system timers.

The document claims that dispatch latencies can be made deterministic and that Sun will (at some future point) be publishing these for each hardware config.

Obviously Solaris is totally unsuitable for applications like embedded weapons control, but is probably suitable for process control, factory automation, etc.

I'd suggest you contact your local Sun sales office and find out when Sun will be publishing the latency figures for the various hardware configurations, and whether they have a later manual.

BTW we use an IPX running 4.1.2 for high-volume collection of weather data on a serial port, and it doesn't miss data too often, but it DOES miss once in a while ...

-- Ed Arnold * NCAR * POB 3000, Boulder, CO 80307-3000 * 303-497-1253(voice) 303-497-{1298,1137}(fax) * era@ncar.ucar.edu [128.117.64.4] * era@ncario.BITNET

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Again, thanks for your help.

--------------------------------------------------------------------- Joerg Theuerkauf /| joerg@wp-470-1.fl.bs.dlr.de / | (FL1A @ DLRVMBS.BITNET) / | Tel. (+49) 531 395 22 91 _____/___|_____ Fax (+49) 531 395 25 50 / / / / / / / / /____/____/____/ | / Deutsche Forschungsanstalt | / fuer Luft- und Raumfahrt ___ | / Flughafen (o o) |/ D 3300 Braunschweig ---w--U--w-----------------------------------------------------------



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:46 CDT