1 INFO-VAX	Sun, 17 Jul 2005	Volume 2005 : Issue 396       Contents: Free VAXen (Salinas, CA)D Re: Now the UK Has it's own date in the history of al quiada attacks1 Re: Offshore Telephone Technical support to India 1 Re: Offshore Telephone Technical support to India  Re: Upgrading VMS / Re: VMS process priorities and system processes / Re: VMS process priorities and system processes / Re: VMS process priorities and system processes / Re: VMS process priorities and system processes / Re: VMS process priorities and system processes / RE: VMS process priorities and system processes   F ----------------------------------------------------------------------    Date: 16 Jul 2005 23:51:29 -0700( From: Javier Henderson <javier@KJSL.COM>! Subject: Free VAXen (Salinas, CA) - Message-ID: <86r7dyylfy.fsf@skylane.kjsl.com>   @ I've a few VAXen that are be looking for a good home, all of the( equipment below is in working condition.  @ MicroVAX II in BA23, CPU, 16MB RAM, DELQA, and some random cards  D MicroVAX II in BA123, CPU from a MicroVAX 3600, 16MB RAM, DELQA, and random cards  1 MicroVAX II in BA123, CPU, and I forget the rest.   B VAX 4300/700, CPU, 48MB RAM, TK70, UC08, three drives, and for the, life of me I can't remember the size... 2GB?   DS700   < Some random "90" bridges and the rackmount backplane thingy.  D Pickup only in Salinas, CA, about an hour south of San Jose, CA. TheB stuff is at my hangar at the airport, if someone is looking for an$ excuse to go flying, come on over...  < Drop me an email if you're interested and we'll take it from@ there. It'd be great if you came and pick everything up at once.   -jav   ------------------------------    Date: 17 Jul 2005 09:05:02 -0700$ From: "AEF" <spamsink2001@yahoo.com>M Subject: Re: Now the UK Has it's own date in the history of al quiada attacks C Message-ID: <1121616302.733797.323970@g14g2000cwa.googlegroups.com>    John Smith wrote:  > AEF wrote: > > John Smith wrote:  > >> AEF wrote:  > >>> Michael D. Ober wrote: > >>>>> H > >>>>> My original question really was whether modern higher efficiencyH > >>>>> engine technology and higher gearing results in an optimum speedD > >>>>> greater than the arbitrary 55 mph. The question still stands > >>>>> unanswered.  > >>>>>  > >>>>> Dr. Dweeb  > >>>  > >>> 8 > >>>   http://www.fueleconomy.gov/feg/driveHabits.shtml > >>>  > >>> G > >>> I don't know how the curve was obtained, but it is *an* answer to  > >>> your question. > >>>  > >>>  > >>>> Reality Check > >>>>J > >>>> My 2002 Pontiac Montana AWD gets it's best mpg at 47-52 mph on flatI > >>>> ground with no wind.  This is empirical from watching the overhead J > >>>> instant fuel economy reading.  However, driving any distance in theH > >>>> western United States at this speed is fatal - you fall asleep atJ > >>>> the wheel and kill yourself in the resulting crash.  The reality isI > >>>> that given the same environmental conditions (flat road & no wind, H > >>>> for instance) 70 mph is only 1 mpg lower than 50 mph.  Going overD > >>>> 75 really starts to gulp gas as the wind resistance starts to > >>>> really climb. > >>>>I > >>>> My 1990 Pontiac Transport's most efficient speed was around 50 mph  > >>>> as well.  > >>>> > >>>> Mike Ober.  > >>>  > >>> G > >>> Interesting. Do you know how accurate your "overhead instant fuel G > >>> economy reading" actually is? If it's actually directly measuring I > >>> the rate of gas coming out of the fuel injectors it has a chance to F > >>> be reasonably accurate. If it's just measuing engine vacuum, I'd' > >>> expect it to be only approximate.  > >> > >>E > >> My BMW 540 has both an instantaneous readout gauge calibrated in F > >> litres/100 km and a nav system which also calculates average fuelF > >> economy. My best fuel economy is at about 120 Kph (around 70 mph)B > >> without the airconditioner on - I get about 8.9 litres/100 kmE > >> consumption on premium gas. In the city it's at least 50% worse.  > > J > > Hmmm, you have to go up to 70mph just to get 26.5 miles per gallon?! AC > > Corolla could probably beat that at any speed. If you want good & > > mileage, BMW is not the way to go. > I > Who said I wanted good mileage? I have customers with lots of money :-)     = This just supports my statement that it is odd to have higher E mileage-optimal speeds as a technological goal. And if you don't care C about mileage, why have you bothered to measure it? And how did you G measure it? Are you sure you took all influencing factors into account? B Changes in wind, grade, steadiness of your foot(!)? Personal bias?  C I'm almost surprised that these luxury/power cars even have mileage E meters. Those who are interested in mileage wouldn't even be shopping G for such cars. And I can't see a luxury car shopper saying, "Well, this D one has a mileage meter! I'll get it!" I guess it's just part of the' total "lots of bells and whistles" bit.     I > >> The 540 has a pretty good Cd (coefficient of drag), somewhere around I > >> 0.30 IIRC, where a 1.0 represents something akin to a 4'x8' sheet of  > >> plywood face to the wind. > > I > > That helps overall mileage, but I doubt it changes the fact that drag  > > increases with speed.  > @ > Physics is physics, at least at the Newtonian scale of things.  D At least? Are you saying that non-Newtowian physics isn't or somehow2 might not be physics? That doesn't make any sense.   ------------------------------  % Date: Sun, 17 Jul 2005 22:19:47 +1200 $ From: "Lurker" <nowhere@nothing.com>: Subject: Re: Offshore Telephone Technical support to India3 Message-ID: <xgqCe.1188$PL5.153691@news.xtra.co.nz>   : "JF Mezei" <jfmezei.spamnot@teksavvy.com> wrote in message& news:42D9E025.7E89CF15@teksavvy.com...L > The local squirrels and racoons get fed with peanuts from china. The localI > convenience store has bottled water from the region, as well as bottled  water B > from Fiji.   In New Zealand, I saw bottled water from Qubec. In Australia, I! > saw bottled water from Iceland.   < Well, I haven't seen any water from Iceland (can you give meD the trade-mark name to try it?) But it is true that there are things< like Evian and Perrier bottled waters on supermarket shelves@ (but not in your typical corner grocery since demand isn't much)  0 > Moving a service to India isn't wrong per say.  2 Can some native French speaker please correct that9 "per say"? It surely looks wrong but I'm not an authority  on that :-)    ------------------------------    Date: 17 Jul 2005 05:45:13 -0700$ From: "Ed Wilts" <ewilts@ewilts.org>: Subject: Re: Offshore Telephone Technical support to IndiaC Message-ID: <1121604313.242916.244040@g14g2000cwa.googlegroups.com>   H >The problem is you are not playing on a level field. India provides theG >buildings, the guards, the locked parking lots, and says, here use our C >support.  So it is heavily subsidized to get the business. It's an  >unfair advantage.  E What level playing field?  I happen to live in a town of about 12,000 F in Minnesota where I'm watching Medtronics, one of the world's largestG medical device companies (with a market cap of $63B), negotiate with my C city for tax-free financing to move 3500-6000 people into our city. F They won't be paying for road improvements either - there are a lot of> subsidies involved in US business every day.  I happen to likeD Medtronics - they are a very well respected employer in this state -1 but that's just the way business operates *here*.   
    ..../Ed   ------------------------------   Date: 12 Jul 2005 06:52:48 GMT$ From: "Doc." <doc@openvms-rocks.com> Subject: Re: Upgrading VMS7 Message-ID: <Xns96915AD076321dcovmsrox@212.100.160.126>   ) %NEWS-I-NEWMSG, David J Dachtera wrote in # news:42D316FB.E9800AC8@comcast.net     <snip>   > Who certified the auditors?   & Your local mental health professional.   :-)      Doc. --  G OpenVMS:     Eight out of ten hackers prefer *other* operating systems. G http://www.openvms-rocks.com    Deathrow Public-Access OpenVMS Cluster.    ------------------------------  % Date: Sun, 17 Jul 2005 16:04:29 +0200 , From: Albrecht Schlosser <ajs856@tiscali.de>8 Subject: Re: VMS process priorities and system processes, Message-ID: <jhodbd.hms.ln@news.hus-soft.de>   David J Dachtera wrote:  > Albrecht Schlosser wrote:  >  >>Hi,  >>O >>I'm looking for a way to give some application and server processes different 
 >>priorities:  >>! >>highest:  some server processes ' >>high   :  some other server processes ( >>medium :  some communication processes  >>low    :  some other processes >>lowest :  client processes >>M >>This would result in a process priority "window" of 5 different priorities. R >>These priorities should all be higher than normal (interactive) user priorities,I >>which are by default 4. Batch priorities are still lower (from 1 to 3).  >  > H > Well, remember that if one of these processes goes "run away", you mayI > not be able to login and stop it if it's running at higher than DEFPRI, J > and if you're already logged, you may not be able to stop it expediently9 > if it's running at higher than your process's priority.   O Yes, and that's why I didn't simply do it but asked you and others about their   opinions and suggestions.   H > Remember also that this affects CPU scheduling only, not I/O priority.I > CPU intense processes will get more CPU time (at the expense of others) I > as you raise their base priority. I/O intense processes will see little / > gain or detriment from altered base priority.   O Well, unless these (I/O bound) processes have a lower priority than others and  P do not even get that little amount of CPU time that they need to start the next  I/O in a reasonable time.   F > Try to find some other approach first. This will quickly turn into a+ > monster, and it WILL kill you eventually.   K Thanks for your warnings, I appreciate this and all replies from the other  M posters. I'm very cautious with this, because I can test these modifications  M only in a life production system (there is no test system with enough system  ' load to test these performance issues).   J Here are some more informations why I wanted to classify the processes by Q priority: the system is used only for one application, but there are many users.  O These users connect to the system with client programs via TCPIP services. The  K server processes that are started this way should have the lowest priority  , because they are the user interface servers.  P The highest priority process is a kind of database server that is I/O bound and N uses little CPU, but should get the CPU ASAP, because it serves (potentially) Q all other processes in the application system. (You may say that's a bad design,  < but it is as it is and can't be changed in the near future).  Q The next priority class (high) would be one or more server processes that modify  O in memory tables and do less I/O. They must be fast, but have a lower priority   than the database server.   M The medium priority class would be for communication processes that exchange  I data with external devices (e.g. laboratory devices) that need a certain  P reaction speed. These processes are optimized in a way that they can get almost Q all information from memory tables, and thus they are also CPU bound, but not as  0 important as the high priority server processes.  P The low priority class would be processes that exchange data with other systems L in a batch manner - low priority, but they must do their jobs without being 0 blocked by the "lowest" priority user processes.  O Below these priorities there should be the "maintenance" user processes. If an  L interactive user does a maintenance task, then the normal user applications $ shouldn't be interrupted or blocked.  J So far the theory. The problem with priorities arised since more and more L processes use our big memory tables and thus are mainly CPU bound. I/O as a O method of (CPU) synchronization doesn't work in such an environment as well as  $ in a "normal" multiuser environment.  5 Thanks for your time reading this and all suggestions    Albrecht   ------------------------------  % Date: Sun, 17 Jul 2005 16:05:32 +0200 , From: Albrecht Schlosser <ajs856@tiscali.de>8 Subject: Re: VMS process priorities and system processes, Message-ID: <gjodbd.hms.ln@news.hus-soft.de>   Jeff Chimene wrote:  > Albrecht Schlosser wrote:  >  >>Hi,  >>E >>I'm looking for a way to give some application and server processes  >>different priorities:  >>! >>highest:  some server processes ' >>high   :  some other server processes ( >>medium :  some communication processes  >>low    :  some other processes >>lowest :  client processes >>A >>This would result in a process priority "window" of 5 different ? >>priorities. These priorities should all be higher than normal I >>(interactive) user priorities, which are by default 4. Batch priorities   >>are still lower (from 1 to 3). >  > <snip> >  > Hi Albrecht, > C > Have you found a hard requirement for this design? Properly coded = > applications will work very well w/ normal VMS scheduling.    N Not really a "hard requirement", as the system is working as expected most of L the time. Many of our processes tend to be more and more CPU bound, because Q we're using big memory tables for active data, and thus most of the "I/O" can be  2 done from memory tables and/or RMS global buffers.  
 > The idea is K > that most processes will be in some wait state until they want attention.   O Yes, the real problem is when one of these processes "runs away" in a CPU loop  Q or when a user does something CPU intense with a higher priority than one of the  P critical server processes. We found out recently that a critical server process P ran with priority 3, and user processes with priority 4 could nearly block this P process. This was a configuration problem, and I corrected it, but this lead to M the question: how can it be done even better and more robust without getting   other problems?    Albrecht   ------------------------------  % Date: Sun, 17 Jul 2005 16:04:48 +0200 , From: Albrecht Schlosser <ajs856@tiscali.de>8 Subject: Re: VMS process priorities and system processes, Message-ID: <5iodbd.hms.ln@news.hus-soft.de>   Chris Scheers wrote: > Albrecht Schlosser wrote:  >  >> Hi, >>G >> I'm looking for a way to give some application and server processes   >> different priorities: >>" >> highest:  some server processes( >> high   :  some other server processes) >> medium :  some communication processes ! >> low    :  some other processes  >> lowest :  client processes   
 -- snip --   > B > Keep in mind that if you don't use realtime priorities, various F > activities (primarilly I/O completions) will cause various priority ( > boosts, changing your active priority. > F > Priority is used as a mechanism to choose which process to run when K > there are multiple processes eligible to run.  Do your processes spend a  7 > lot of time eligible to run, i.e, are they CPU bound?   K Yes, some of them are CPU bound, but there is at least one critical server  O process that does disk I/O (RMS indexed files) and that is a server for almost  O all other applications, something like a database server process. This process  P uses very little CPU time, because it's I/O bound, but should get the CPU ASAP, ( when needed - hence priority: "highest".  A > Would some other synchronization mechanism be more appropriate?   Q No, I don't think so, because most of the processes are independant from others,   it's a multi user system ;-)  I > If you have processes which may become stuck in a CPU bound loop, they  + > should not be run higher than priority 4.   P Who knows? In an ideal world there are no such processes, but what I want is to & limit the ramifications if it happens.   Albrecht   ------------------------------  % Date: Sun, 17 Jul 2005 16:06:35 +0200 , From: Albrecht Schlosser <ajs856@tiscali.de>8 Subject: Re: VMS process priorities and system processes, Message-ID: <glodbd.hms.ln@news.hus-soft.de>   Main, Kerry wrote: >  >>-----Original Message----- >>From: Albrecht Schlosser >>Sent: July 15, 2005 6:55 AM  >>To: Info-VAX@Mvb.Saic.Com 6 >>Subject: VMS process priorities and system processes >> >>Hi,  >>< >>I'm looking for a way to give some application and server  >>processes different 
 >>priorities:  >>! >>highest:  some server processes ' >>high   :  some other server processes ( >>medium :  some communication processes  >>low    :  some other processes >>lowest :  client processes >>8 >>This would result in a process priority "window" of 5  >>different priorities.   
 -- snip --  9 >>Additional background information: Of course, the high   >>priority server processes = >>are tested well and stable, but some other processes might   >>come into a CPU loop  @ >>or similar (e.g. once there was a process with a read timeout  >>of 0 seconds, = >>which came from a (missing) user input timeout value :-( ).  >>@ >>In such cases a system manager should still be able to log in  >>and suspend/stop  < >>such bad behaving processes. And network activities (many  >>processes connect via = >>TCPIP services) should never be blocked by other processes.  >> >>Any suggestions what to do?  >>	 >>Regards  >>
 >>Albrecht >> >  > D > It would appear that you want to ensure some processes of the sameE > priority do not impact other processese e.g. process runaway stuff.   L There are two goals: First of all the system should operate with reasonable I performance for all users under normal conditions. The second goal is to  Q minimize the impact of misconfigured or bad behaving ("run away") processes that  0 could otherwise block critical server processes.  E > As others have stated, you can get into trouble by purposely mixing $ > priorities on specific priorities.  ( I don't really understand this sentence.  G > As a suggestion, I would suggest you consider using the OpenVMS class  > scheduler to do this.  >  > Reference: > Sysman> Help class > J > http://h71000.www7.hp.com/doc/732FINAL/6048/6048pro_005.html#startsubcom
 > mand_431F > "The class scheduler provides the ability to limit the amount of CPUC > time that a system's users receive by placing users in scheduling H > classes. Each class is assigned a percentage of the overall system CPUC > time. As the system runs, the combined set of users in a class is D > limited to the percentage of CPU execution time allocated to their	 > class.   > H > Users might get some additional CPU time if the qualifier /WINDFALL isF > enabled for their scheduling class. Enabling the qualifier /WINDFALLF > allows the system to give a small amount of CPU time to a schedulingJ > class when the scheduling class's allotted time has been depleted, but a > free CPU is available."   L Interesting suggestion. This didn't come to my mind, though I read about it ? somewhere some time ago. I'll check if it can be of use for us.   	 > Regards  >  > Kerry Main > Senior Consultant  > HP Services Canada > Voice: 613-592-4660  > Fax: 613-591-4477  > kerryDOTmainAThpDOTcom > (remove the DOT's and AT)  > 6 > OpenVMS - the secure, multi-site OS that just works.   I second this ;-)    Albrecht   ------------------------------  % Date: Sun, 17 Jul 2005 12:21:16 -0400 - From: JF Mezei <jfmezei.spamnot@teksavvy.com> 8 Subject: Re: VMS process priorities and system processes, Message-ID: <42DA857B.F990C5EF@teksavvy.com>  N If your system runs just one application, then why not make your batch defaultM priority 0, and then you can use 1 2 3 for lower priority processes and 4 for 0 your highest and interactrive system management.  K (priorioty 0 is akin to the iddle process, although I am not sure if it has = any special coding in VMS or if it is just another priority).    ------------------------------  % Date: Sun, 17 Jul 2005 12:42:47 -0400 ' From: "Main, Kerry" <kerry.main@hp.com> 8 Subject: RE: VMS process priorities and system processesR Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB651349@tayexc19.americas.cpqcorp.net>   > -----Original Message-----8 > From: Albrecht Schlosser [mailto:ajs856@tiscali.de]=20 > Sent: July 17, 2005 10:07 AM > To: Info-VAX@Mvb.Saic.Com : > Subject: Re: VMS process priorities and system processes >=20   [snip..]  G > > As others have stated, you can get into trouble by purposely mixing & > > priorities on specific priorities. >=20* > I don't really understand this sentence. >=20  E Oops, my apologies - slippery fingers. The statement should have read C "you can get into trouble by manually mixing priorities on specific  processes."   E Fwiw, whenever I give talks about virtualization and/or consolidation E (on any platform), workload management (WLM - not to be confused with A the prod on HP-UX) is one of the main key's to making application H stacking work. Workload management is certainly not a new concept and inH fact, the mainframe folks live and breathe this stuff all the time. Some6 may know WLM as another term ie. "class scheduler".=20  G In a nutshell, you decide what processes should be part of a scheduling F class and then decide what service level objectives (SLO) are for thatE group i.e. "it is low priority, so that group of processes should not F take more than 15% of the overall cpu resources between x and y hrs ofH the day" while another group might be "medium priority, hence this group  can take up to 35% of the CPU.."  @ While OpenVMS's priority and multi-tasking design has made it anG extremely stable multi-tasking OS, it is not immune from the occasional @ process doing something silly (ask those who implemented earlierG versions of WEBES diag stuff), so WLM or the class scheduler is one way C to further minimize the risk of the "single process doing something  silly" problem from occurring.  G And the class scheduling components are part of the OS, so there are no  additional charges.    Regards   
 Kerry Main Senior Consultant  HP Services Canada Voice: 613-592-4660  Fax: 613-591-4477  kerryDOTmainAThpDOTcom (remove the DOT's and AT)=20  4 OpenVMS - the secure, multi-site OS that just works.   ------------------------------   End of INFO-VAX 2005.396 ************************