Sun-Spots-Request@RICE.EDU (William LeFebvre) (08/15/88)
SUN-SPOTS DIGEST Sunday, 14 August 1988 Volume 6 : Issue 181 Today's Topics: Re: Photographic Output Devices (2) Re: problems with forwarding mail Re: ups on sun Re: Suntools startup bug/feature Re: Whither 68030 Announcing top, version 2.5 memory on 4/110 Xylogics 753 simple timing tests How do I get internal *and* external SCSI on a 3/260 w/ 3.4? Send contributions to: sun-spots@rice.edu Send subscription add/delete requests to: sun-spots-request@rice.edu Bitnet readers can subscribe directly with the CMS command: TELL LISTSERV AT RICE SUBSCRIBE SUNSPOTS My Full Name Recent backissues are available via anonymous FTP from "titan.rice.edu". For volume X, issue Y, "get sun-spots/vXnY". They are also accessible through the archive server: mail the request "send sun-spots vXnY" to "archive-server@rice.edu" or mail the word "help" to the same address for more information. ---------------------------------------------------------------------- Date: Mon, 8 Aug 88 12:29:37 EDT From: gfr%wolfgang@gateway.mitre.org (Glenn Roberts) Subject: Re: Photographic Output Devices (1) Reference: v6n172 > I am interested in purchasing a photographic output device for our Sun > 3/110C. The type of device we are looking at is something like the Matrix > recorder... but we aren't interested in a mega-bucks super-camera. We shoot directly from the screen and have had excellent results. We use a 35mm w/ 85mm focal length mounted on a sturdy tripod. With room completely dark and Ektachrome 200 daylight I use an exposure of f8 at 1/2 second (I presume you have the 19" screen; set the brightness and contrast to max). We've also shot off a matrix camera and most people can't tell the differerence between the two methods. The price is right!! (P.S.: Greetings to all my friends at Rensselaer!) - Glenn Roberts (BS '76, M.Eng '78, Ph.D. '80) MITRE Corp., McLean VA (703) 883-6820 gfr%wolfgang@gateway.mitre.org ------------------------------ Date: Mon, 8 Aug 88 09:35:24 PDT From: chaos%gojira.Berkeley.EDU@jade.berkeley.edu Subject: Re: Photographic Output Devices (2) Reference: v6n172 David, I just photograph off the (color) screen with a standard 35mm camera on a tripod. (Ektachrome 160 Daylight; exposures of 1 to 2 seconds) Very inexpensive. And quick if you have 1 hour processing nearby. If you are photographing a monochrome screen with color film, filters are available to remove the color distortion. For animation I use a Bolex 16mm camera pointed at the screen for film output and a Lyon-Lamb VCR controller + JVC CR 850 3/4" VCR + Datacube frame buffer for video. By the way, if you get any information on the several $1K output boxes. Please forward this on. You realize that most of these typically required RS-170 RGB video source (!= Sun video) and so additional $. Jim Crutchfield Physics, UCB (415) 642-1287 ------------------------------ Date: 8 Aug 88 16:25:25 GMT From: woods@handies.ucar.edu (Greg Woods) Subject: Re: problems with forwarding mail In reply to the person who complained that .forward files only work on the server machine when the home directory is protected: the reason for this is obvious if you think about it for a minute. sendmail(8) runs as root, so it has no problem accessing the .forward file on the server. However on the workstations it is trying to access the file through NFS and therefore ends up with the permissions of "nobody" and hence is denied access. Anyone who is so paranoid that they have to have their home directory mode 700 is going to have other problems using public facilities such as electronic mail. My recommendation is to make the home directory mode 711, and protect the files and directories within it if necessary. One other possibility: sendmail will ignore the .forward file if it is not owned by the user whose mail is being forwarded. Even if it is owned by "root". You should check to make sure this is the case. --Greg (postmaster@ncar.ucar.edu) ------------------------------ Date: Mon, 8 Aug 88 10:27:29 EDT From: eugene@gatech.edu (Eugene Galiano) Subject: Re: ups on sun i don't know of any direct way for the sun to detect a power outage, but the way i cludged it is the following: connect a terminal (or some other device like a printer or whatever) to one of the rs-232 portss on the sun and plug it to the wall. Have a dummy user logged in on the terminal (or some process running) when the power goes out, the terminal powers down and the user is logged off, at this point start up a shutdown script that will shutdown the system. To detect if the user is logged on simply use finger or who. eugene ------------------------------ Date: Mon, 8 Aug 88 07:48:53 EDT From: reg%lti.UUCP@bu-it.bu.edu (Rick Genter x18) Subject: Re: Suntools startup bug/feature Reference: v6n166 In Sun-Spots Digest, v6n166, munnari!yabbie.rmit.oz.au!rcodi@uunet.uu.net (Ian Donaldson) writes: > Presumably suntools is being smart and using the error code returned from > the open (which should be EPERM) to loop and try again. Why it doesn't > try the next available pty I have no idea. Maybe it thinks that any error > code other than EBUSY means there are no free pty's?? Not having access to Sun sources, I can only hazard a guess, but is it possible that shelltool loops on an error opening the pty because of the possibility of an NFS error? I agree that looping infinitely is an unacceptable behavior, but it is certainly better than SCCS's behavior, which is to abort whatever it's doing if an "NFS server not responding; still trying" condition occurs. - reg ------------------------------ Date: Mon, 8 Aug 88 10:58:58 edt From: petsche@demon.siemens.com (Tom Petsche) Subject: Re: Whither 68030 Reference: v6n167 >Both HP and Apollo have recently announced 7-MIP 33 MHz 68030 >workstations. I've heard not a single rumor about any possible 68030 Sun >workstation. As I recall a conversation with out neighborhood Sun salesman, a 68030 machine, rated at 5 to 7 MIPS is due the last quarter of 88 (or maybe first quarter of 89). A new Sun4, rated at 20-25 MIPS is due Dec 88. tom ------------------------------ Date: Fri, 12 Aug 88 16:25:38 CDT From: William LeFebvre <phil@Rice.edu> Subject: Announcing top, version 2.5 Thanks to the tireless efforts of a former roommate of mine, Scott Alexander, I now have a version of "top" that works under SunOS 4.0. This extra functionality has been embedded in "top" version 2.5. For those of you who don't know, "top" is a top-processes display for Unix, inspired by the "top 10" display of VMS. It displays the top CPU-using processes on the system and updates this information periodically. It also displays some overall system information (number of processes, idle time, memory usage, etc.). Version 2.5 is now available in the Rice archives. It just so happens that the same mecahnisms used for accessing the Sun-Spots archives can be used to access the generic "public" area at Rice. Top is contained in four shar files underneath "public": "top.shar.01" (31152 bytes), "top.shar.02" (29241 bytes), "top.shar.03" (29282 bytes), and "top.shar.04" (27080 bytes). These files can be retrieved via anonymous FTP from the host "titan.rice.edu" or via the archive server with a request like "send public top.shar.01". For more information about the archive server, send a mail message containing the word "help" to the address "archive-server@rice.edu". I plan on sending this version to the unix-sources list soon. William LeFebvre Department of Computer Science Rice University <phil@Rice.edu> ------------------------------ Date: Mon, 8 Aug 88 10:35:21 EDT From: gotham!marine!fish@sun.com (Lorenz Fish) Subject: memory on 4/110 The 4/110 uses the Static Column behavior of the SCRAMs to simulate a cache - there is not a real cache per se. Once a particular column is addressed in a SCRAM row addresses within that column can be accessed without re-strobing the column address. In most cases that cuts access time from 3 or 4 cycles to one cycle (in the case of a "hit"). There is a custom gate array on the CPU board which keeps track of the last accessed column addresses and which heads off column addressing if there is a "hit". Simultaneously the column address is being prepared (by the MMU) so that in the case of a "miss" no time has been lost doing the column address comparison. There are many configurations possible utilizing 256K SCRAM SIMs and 1Meg SCRAM SIMs, not all of which are supported (avoids market/sales/service nightmares). Last I knew you could order the board with 8, 16, 20, or 32 Meg but delivery is severely affected by availability of the SIMs. Lorenz Fish Technical Analysis Group Marine Midland Bank Manhattan (former diagnostic software member of Sun 4/xxx bringup teams) ------------------------------ Date: Sun, 7 Aug 88 22:12:18 MDT From: hwe@lanl.gov (Skip Egdorf) Subject: Xylogics 753 simple timing tests I have just completed some simple performance tests of a new Xylogics 753 (supposed to be identical to the pseudo-mythical Sun 7053), a Ciprico Rimfire 3200, and a Sun Xylogics 451 on a Sun 3/280S with Fuji 2361As. I added a CDC 9772 on another 3/260 driven by a Ciprico rf3200 and some Wren-IVs on SCSI controllers just to broaden the perspective a bit. I already own some other xy451s (from Sun) and a couple of rf3200s. The main purpose here was to see how well the 753 would work with the 'xd' driver in SunOS 4.0 (It was fine: Pure plug-and-play. Unlike the previous xy751 and the xy752 that never did work, the 753 seems to be a fine product.), as well as see how it compares to other controllers with which I am already familier. By the way, I don't work for, nor own stock in Xylogics, Ciprico, Sun, CDC, or Fuji. These are my own informal tests, and do not constitute any sort of endorsement (or criticism) of any product by Los Alamos National Laboratory, the DOE, the University of California, or any other organization whose lawyer's may read this. Criticism, or additional interpretation may be either e-mailed to hwe@lanl.gov, or posted, as you think best. Thanks to Dale Carstensen of LANL C-3 for the loan of the 753. Thanks to Don Wood of Frontier Systems in Albequerque for the loan of the new rf3200 used in these tests. The first impression of the controller is nice. It is all surface-mount technology, and the SMD cable connectors are a new mini-D style that includes a locking clamp, and a built-in ground strap. You WILL need the special cables. The systems were: 1. 3/280S, xy451, Sun/Fuji 2361A, SunOS 4.0, Sun xy device driver. 2. 3/280S, xy753, Sun/Fuji 2361A, SunOS 4.0, Sun xd device driver. 3. 3/280S, rf3200, Sun/Fuji 2361A, SunOS 4.0, Ciprico rf driver. 4. 3/260S, rf3200, CDC 9772-850, SunOS 3.5, Ciprico rf driver. 5. 3/260S, SCSI-2, CDC Wren-IV, SunOS 3.5, Sun sc device driver. 6. 3/60, SCSI-3, CDC Wren-IV, SunOS 4.0, Sun si device driver. The rf3200 is a rev-C. The xy-753 is a '-101' (see below). The Ciprico device driver is version 1.10 (An earlier poster to sun-spots mentioned trouble running rfutil under 4.0. Mine worked fine.) The tests were: read test: 'dd' of a raw partition of 16080 sectors to /dev/null. On the CDC Wrens and the 9772, a 'bs=' was used to ensure the same amount of data transferred in each test. write test: a C program was used to write 5 Mbytes of data to a raw disk partition. NOTE that the read and write tests used different total sizes, and therefore should not be directly compared with each other. Read was about 8 MBytes. Write was 5 MBytes. Each test was performed with block sizes of 1 kbyte: chosen as a worst case. Suns will practically never use a block size this small. 8 kbyte: chosen as typical of Sun Disk IO. 63 kbyte: Chosen as a best case. The '126b' of dd fame is chosen for the same reason. Again, unrealistic for normal use. The tests did not prod the Sun file system at all. Everything was pure bandwidth to the raw disk. The tests were run on only a single drive on each controller. The newer controllers are supposed to handle multiple drives better than the xy451 (or the 'sc' SCSI), but these tests have nothing to do with that issue. These tests will show little about the effect of any on-board cache. Each test was run several times. The results did not vary. Now for the numbers: all times in minutes:seconds from the csh 'time' command. test 3/280 3/280 3/280 3/260 3/260 3/60 4.0 4.0 4.0 3.5 3.5 4.0 xy/2361 xd/2361 rf/2361 rf/9772 sc/Wren si/Wren read 1k 2:18 :35 :37 :36 2:20 2:21 read 8k :17 :06 :07 :06 :22 :19 read 126b :08 :04 :04 :04 :12 :12 write 1k 1:28 1:28 1:28 1:28 1:29 1:29 write 8k :13 :13 :13 :14 :14 :14 write 126b :06 :04 :04 :04 :07 :07 Now for my interpretations. (others please feel free to post or mail other interpretations.) 1. The xy753 shows about a 3:1 speedup over the xy451 in the read tests for typical Sun block sizes. 2. I am continuously amazed at how cost-effective a 3/60 with a couple of Wren-IVs is. The scsi performance with the 1.2 MB/sec drives is about the same as a 3/280 xy451 with Eagles. The 753/7053 will fix this imbalance. 3. The write test was surprising, as the performance gains of the 32-bit controllers only showed up on the large block tests. It seems that controller performance is much less a factor on the write test than the read test. (Anyone know why??) 4. The speed difference between the 3-MByte/second CDC 9772 and the 2.4-MByte/second Fuji 2361 didn't show up. (These tests did NO test of the seek performance) 5. The XY-753 seems to be the winner by a hair. Then again it is a year newer than the rf3200. The ability to plug-and-play is a lot more important than the disk performance. Finally some notes: The xy-753 comes in 2 versions (part-numbers from Xylogics). the xy753-101 runs 2.4 MB/second drives. the xy753-103 runs 3 MB/second drives. Xylogics has been shipping about 1/2 and 1/2, so check the label on the back of the board to see which you have. The only difference is ECL drivers on the 3 MB/sec version. I have a good rumor that the Sun 7053 will be available in only the 2.4 MB/sec version. This may change by the time the 7053 is officially announced, but then the 4.0 driver, the v2.6 boot proms, and the section 4 man page must make the 7053 the most announced unannounced product in history. For those who don't know, the xy753 is a standard 2-high VME card. The 7053 will be the same chips and etch, but on a 3-high Sun-sized board to avoid the need for a form-factor adapter. The 7053 will be available ONLY from Sun. I did a full load of the 4.0 tapes on the 753, including format from munix, load the miniroot, and build a real vmunix all on the 753. I also did an online format from a system running on the other 2361 and the 451. EVERYTHING worked properly thinking that there was an 'xd' disk. The 753 is a true plug and play. Skip Egdorf hwe@lanl.gov ------------------------------ Date: Mon, 8 Aug 88 11:05:38 edt From: petsche@demon.siemens.com (Tom Petsche) Subject: How do I get internal *and* external SCSI on a 3/260 w/ 3.4? We recently got an Exebyte 8mm video tape backup unit which is performing nicely (From perfect byte, we received it within two days of odering it). Now the problem. The unit plugs into a SCSI bus. We currently have it plugged into a 3/60 diskless client and do dumps over ethernet. I'd like to plug the 8mm tape into our 3/260 server, but the server has an internal SCSI tape and disk. The easiest solution would probably be to get another SCSI controller, but 3.X can't handle two SCSI controllers, but I'm not ready to upgrade to 4.X yet. The question: Does anybody know where I can get a SCSI controller for a 3/260 that can handle both internal and external SCSI devices? Thomas Petsche petsche@siemens.com petsche@ivy.princeton.edu ------------------------------ End of SUN-Spots Digest ***********************