1 INFO-VAX	Sun, 13 Nov 2005	Volume 2005 : Issue 634       Contents:! Re: A question about autocomplete ? RE: failSAFE IP - Looks good! Who's using it? (and why/why-not) . RE: FW: OT: Microsoft drop more Itanic supportG RE: Will Rich Marcello either come clean or do some proper marketingand G RE: Will Rich Marcello either come clean or do some proper marketingand   F ----------------------------------------------------------------------  % Date: Sun, 13 Nov 2005 11:20:17 +0000 # From: issinoho <issinoho@gmail.com> * Subject: Re: A question about autocomplete4 Message-ID: <dl77lh$30c$1$830fa7a5@news.demon.co.uk>   Peter Weaver wrote: 2 > "issinoho" <issinoho@gmail.com> wrote in message0 > news:dl4n4r$cgj$1$8300dec7@news.demon.co.uk...J >> I've always felt that one of the biggest failings with DCL was the lackJ >> of a Unix-style autocompletion of filepaths when TAB is hit. This wouldB >> be a fantastic addition, however if I remember rightly an olderI >> discussion around this offered some fundamental reasons why this would  >> never work under DCL. >> ... > I > Does nobody here keep an eye things coming out of the OpenOffice group?  > % > http://www.oooovms.dyndns.org/auto/  >  >   # Not *really* the same though is it?    ------------------------------  % Date: Sun, 13 Nov 2005 09:11:25 -0500 ' From: "Main, Kerry" <Kerry.Main@hp.com> H Subject: RE: failSAFE IP - Looks good! Who's using it? (and why/why-not)R Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB70CC40@tayexc19.americas.cpqcorp.net>   > -----Original Message-----9 > From: JF Mezei [mailto:jfmezei.spamnot@teksavvy.com]=20 " > Sent: November 13, 2005 12:59 AM > To: Info-VAX@Mvb.Saic.Com @ > Subject: Re: failSAFE IP - Looks good! Who's using it? (and=20 > why/why-not) >=20 > "Main, Kerry" wrote:> > > DNS Alias with Load Broker provides system availability=20 > features and the> > > IP Failover provides capabilities to increase the link =20 > availability andH > > transmit throughput. The Load broker actually uses host based server< > > busy data to determine which nodes should get the new=20 > connections, so + > > this is a better way of load balancing.  >=20 >=20@ > Out of curiosity, if load balancing is done via DNS tricks,=20 > what sort of* > TTL values are used for DNS entries ?=20 >=20F > For instance, if you try to download a standard MS Frontage designedA > page, you will have a gazillion HTTP transactions. Once your=20 
 > client gets A > the IP for WWW.CHOCOLATE.COM, it will reused that IP for the=20 
 > duration of H > the transaction and if you skip to the next page, it will re-use it asF > well, unless you have an abnormally low TTL of just a second or two. >=20  G Yep, the TTL value will be a trade-off that one would need to make i.e. B more load on DNS server vs how quickly the system load changes areH adapted to and how soon a client will be re-directed to a new node after2 the original node failed or was taken off line.=20  F More info on load broker is available at: (specifically section 6.1.3)H http://h71000.www7.hp.com/doc/73final/6526/6526pro_contents_001.html#toc
 _chapter_6  G This load broker technique is how one can shutdown servers in a cluster G with zero application availability impact. Simply set flag (as I recall A setting logins=3D0 will do this) and allow current connections to @ complete, but direct all new connections to other servers in theG cluster. When all connections are complete, that server can be shutdown + - and end users do not need to be informed.   C [As I recall, when the load broker detects logins=3D0 on a node, it H removes that specific node address from the TCPIP cluster alias options]  C There are some batch things to consider, but they can be planned as  well.   G Note - the failSAFE IP will improve transmit throughput, so if you have F something like an FTP server, then this might be something to consider# as well as the availability aspect.      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.   ------------------------------  % Date: Sun, 13 Nov 2005 12:21:15 -0500 ' From: "Main, Kerry" <Kerry.Main@hp.com> 7 Subject: RE: FW: OT: Microsoft drop more Itanic support R Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB70CC46@tayexc19.americas.cpqcorp.net>   > -----Original Message-----4 > From: Bill Todd [mailto:billtodd@metrocast.net]=20! > Sent: November 10, 2005 9:50 PM  > To: Info-VAX@Mvb.Saic.Com 9 > Subject: Re: FW: OT: Microsoft drop more Itanic support  >=20 > Main, Kerry wrote: > >>-----Original Message-----5 > >>From: Dave Froble [mailto:davef@tsoft-inc.com]=20 " > >>Sent: November 9, 2005 9:39 PM > >>To: Info-VAX@Mvb.Saic.Com ; > >>Subject: Re: FW: OT: Microsoft drop more Itanic support  > >>  	 [snip ..]   ; > So while it is indeed true that neither x86 nor Itanic=20  > *currently* has a=20I > great deal of influence in the upper reaches of the market served by=20 @ > POWER5 and SPARC, x86 has far better prospects for changing=20 > its position=20 + > significantly there than Itanic seems to.  >=20  E You are only looking at the Mhz and speeds-n-feeds trenches. Raise up  your sights a bit.  F Keep in mind the 800lb gorilla problem that the majority of all med toE large Customers today are facing i.e. way to many distributed servers D that are grossly under utilized. Wintel systems are typically 10-25%7 busy and UNIX systems are 15-30% busy in prime time.=20   H So, IT managers are not that interested in how these new bits-n-bites inH the new servers can make their utilization rates go even lower. They areH much more interested in solutions that can lower their costs and improve0 the level of service offered to their end users.  G Windows and Linux are not going to replace the application environments H currently used by AIX, Solaris, OpenVMS, MVS (or Z/OS as I believe it isH called now). Will Wintel/Linux potentially impact some of these markets? Absolutely.=20  G However, these environments are also susceptible to downtime and having C to QA/test monthly Wintel/Linux security patches with most of their H important applications is not going to happen anytime soon. In addition,F as I mentioned before, the one-app, one server model is under a lot of scrutiny these days.  F My point is that Intel recognizes the chip is only a small part of theC overall solution and they need something to compete with Power5 and  SPARC.    4 >   X86 and x86-64 sweet spot is still very much 2-4 > > cpu systems. >=20G > So, one might note, is Itanic's:  commercial performance falls off=20 : > significantly above the 4-socket system level (which,=20 > incidentally, is=20 A > the only area where it can stay within anything like hailing=20  > distance of=20
 > POWER5). >=20= >   That is also limited by applications on Windows and Linux B > > to scale reliably above 5+ cpu's (yes, there are some app's=20 > that do a > > > bit better, but not many). Good example is MS Exchange.=20 >=20A > So what?  Are you acquainted with any Exchange installations=20  > that can't=20 I > run more than comfortably on 5 or fewer CPUs?  If so, exactly how to=20 A > they manage today (and how would Itanic help them)?  If not,=20  > what's your=20 > point?   You are joking - right?   F With Exchange servers typically handling only a few thousand users perH server, very limited multi-site DR capabilities, how many servers do youC think it requires to handle large corporations or institutions with H 100K+ employees (include DR in those calculations). The issue is not cpu4 speed but rather transaction logs, recovery time etc  D The point is that there are mail packages that scale much better andF have much better DR capabilities than Exchange, but they typically areC not run on Wintel or Linux if they do not want to deal with monthly - security patches and need a good DR solution.    >=20I > IBM's Hurricane x86 system is partitionable, you know.  Applications=20 G > that don't scale well on x86 won't scale any better on Itanic, and=20 J > applications that do scale well (e.g., SQL Server) will do so just as=20& > well on x86 as they would on Itanic. >=20 > >=20B > > Hence, Intel needs a higher end offering to compete against=20 > those Power5 > > and SPARC platforms. >=20F > It's looking more and more like Xeon just became Intel's high-end=20 > offering of choice.  >=20  D You are only thinking of hardware bits-n-bytes. Real world solutions@ need to consider much more than how many Mhz a CPU or FSB is.=20  D Remember the typical CIO has a huge number of x86 servers running atG only 10-25% (some in consolidation business might say 5-15%) busy right C now and they were all bought over the last number of years with the @ latest feeds-n-speed, one app, many servers type discussions. InE retrospect, given these utilization numbers, do you think these CIO's D think it was worth buying the latest-n-greatest X86 servers over the last few years? =20   F Perhaps a "middle of the road" config that allowed them to standardize" might have been a better strategy?  F Good example - consolidation RFP for one large company stated they hadD 11,000 servers - 8,000 Wintel. They stated the Wintel average server; utilization in peak time was estimated to be approx 10%.=20   H This type of industry problem is certainly not being lost by Intel.  TheG one applic, many server (dev/test/QA/prod) model is in for some serious $ re-thinking over the next few years.   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.   ------------------------------  % Date: Sun, 13 Nov 2005 09:41:45 -0500 ' From: "Main, Kerry" <Kerry.Main@hp.com> P Subject: RE: Will Rich Marcello either come clean or do some proper marketingandR Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB70CC41@tayexc19.americas.cpqcorp.net>   > -----Original Message-----= > From: Richard Maher [mailto:maher_rj@hotspamnotmail.com]=20 " > Sent: November 12, 2005 11:38 PM > To: Info-VAX@Mvb.Saic.Com A > Subject: Re: Will Rich Marcello either come clean or do some=20  > proper marketingand  >=20 > Hi Kerry,  >=20B > It doesn't appear that you're listening to me anymore but for=20 > the benefit ofH > myself (if no one else) I can't let the following complete bollocks go > unchallenged: -  >=20A > >  In OpenVMS terms, that would include using technologies such D > > as RTR which, similar to HW fault tolerance, ensures a committed > > transaction is never lost. >=20/ > Did you mis-type RTR and were meaning DECdtm?  >=20= > If not were you talking about RTR's store-and-forward or=20  > guaranteed-delivery @ > functionality when you were talking about a "transaction is=20 > never lost"? If 9 > it's the latter then could you please answer me this: -  >=20B > What happens if the first part of the RTR transaction is give=20 > Fred 1 million? > dollars and the second part is take 1 million dollars from=20  > Jack. Now when= > the transaction was started Jack had a million bucks but=20  > since the network ? > link to the remote database holding Jack's account-details=20  > has been down,= > he's transferred that 1M squid to his bookie. So RTR can=20  > stamp up and down = > as much as it wants with its "I want a million dollars!"=20  > (don't we all :-) > > but it's just not gonna get it. Excellent stuff about the=20 > transaction not G > being lost though! The ability to replay stale info is quite a trick.  >=20   Richard,=20   $ I am certainly no guru on RTR, but -  B RTR has the ability to find alternate network routes to complete aA transaction. Hence, if one link or router went down, it will find E another path. And if they are using true 2PC, as you know, either the C transaction completes at both sites or neither of them i.e. it gets 7 rolled back and the application error handler kicks in.   E Keep in mind that in many critical cases, you are talking about using D RTR in conjunction with a multi-site OpenVMS cluster with host basedF volume shadowing, so the data will always be consistent between sites.  E Keep in mind that many stock exchanges use this today, so are you are + stating that you know something they don't?   . Here is a pointer to some additional RTR info:* http://www.hp.com/products1/rtr/index.html) http://www.hp.com/products1/rtr/faqs.html       @ > Look into my eyes, ACID is just-like-sooo-yesterday. It's a=20 > hangover from B > mainframe legacy days that simply cannot model today's web basedB > requirements. . . You are getting sleepy . . . OMX Click uses=20 > it so it must E > work for everyone else. . .You eyelids are getting heavy. . .if web A > transactions didn't work then how does Amazon function? Hey?=20  > Hey? . . .You : > are in your happy place. . .no one can hurt you anymore. >=20 > WAKEUP!!!   @ Wow, I am sure all of the worlds banks will need to hear how theB transaction models they have been using for the last 20+ years has actually not been a good one.   @ As to Amazon, I would have to think that losing a number of $100B transactions for books and some CD's is a whole lot different than2 losing a number of $10M to $100M+ transactions.=20  E Heck, ACID stuff is (and never was) designed to be the right base for H every application. There are always going to be cost-benefit trade-off'sF between complexity, availability and data consistency requirements.=20   [snip..]   Regards,  
 Kerry Main Senior Consultant  HP Services Canada Voice: 613-592-4660  Fax: 613-591-4477  kerryDOTmainAThpDOTcom (remove the DOT's and AT)=20  4 OpenVMS - the secure, multi-site OS that just works.   ------------------------------  % Date: Sun, 13 Nov 2005 09:54:29 -0500 ' From: "Main, Kerry" <Kerry.Main@hp.com> P Subject: RE: Will Rich Marcello either come clean or do some proper marketingandR Message-ID: <FD827B33AB0D9C4E92EACEEFEE2BA2FB70CC42@tayexc19.americas.cpqcorp.net>   > -----Original Message----- > From: Main, Kerry=20! > Sent: November 13, 2005 9:42 AM  > To: Info-VAX@Mvb.Saic.Com A > Subject: RE: Will Rich Marcello either come clean or do some=20  > proper marketingand  >=20 >=20 > > -----Original Message-----? > > From: Richard Maher [mailto:maher_rj@hotspamnotmail.com]=20 $ > > Sent: November 12, 2005 11:38 PM > > To: Info-VAX@Mvb.Saic.Com C > > Subject: Re: Will Rich Marcello either come clean or do some=20  > > proper marketingand  > >=20  % Oops, minor typo I need to correct ..    >=20G > Heck, ACID stuff is (and never was) designed to be the right base for A > every application. There are always going to be cost-benefit=20 
 > trade-off's H > between complexity, availability and data consistency requirements.=20 >=20   Should have read: G Heck, ACID stuff is *not* (and never was) designed to be the right base  for every application.  
 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 2005.634 ************************  cB KG}`o\8̀
OĈF4JIk3=Ʊ6+ZydƐx Ƴ!KVommlm!΀uQDTpJvp`@
>%ooxXjXƋQv,Sd#Yߺm  k
{~|8qh$BZ{blO%$lg,ݗ9)&ƾL)h%:^Qwbzrx#NF<UdM*2ȧVEiՐdD}h&k>rH$%N&:	RPp

Lq>lLF
,ytU86웇BǞW%Ns1$CJ~˞qY&8?,3Ehsh+xWɪvseS*ЛBo=ZSUNكd7WSVm')BUvȪy=;a2Ó]A'^Fx"=Hn%͙A}ף!D6<_98 petX7/iHZe/|EwsG茟
!ڤ[GMGmBi2E9~Ӛc!L@o	k˞-ɾvyХi
̟#ڵ:5LB@L]8vL܎faBtBvHoʐ6Rh74/0NdUuU6i.rPAPcyO7vifXNOyty1}^~:%&E뢏^
4P&^*~yVUtqͥƜeq=HGſAwM[ \aD$eb~xuE~vA0%6?JW96}h'{-[KPX) 26?#hUtI' To`<M