Archive-Date: Fri, 1 Jul 1994 20:50:59 GMT Subject: Remote Script/Command file execution Message-ID: From: epope@netcom.com (Erik Pope) Date: Fri, 1 Jul 1994 20:50:59 GMT I am setting up a system which requires events in an application running under VMS to run scripts/programs unser unix and the reverse as well. I have discovered that this is aparently an non-trivial task. Does anyone know of a method of tools which make remote execution across platforms like these doable? epope@netcom.com or tpopee@tus@ssi1.com ================================================================================ Archive-Date: XXX, 5 Jul 1994 14:20:50 GMT Subject: Re: Remote Script/Command file execution Message-ID: <2vbq82$51u@nntpd.lkg.dec.com> From: huston@icecub.ako.dec.com (Steve Huston) Date: 5 Jul 1994 14:20:50 GMT Reply-To: huston@icecub.ako.dec.com References: In article , epope@netcom.com (Erik Pope) writes: >I am setting up a system which requires events in an application running >under VMS to run scripts/programs unser unix and the reverse as well. .. >Does anyone know of a method of tools which make remote execution across >platforms like these doable? The 'r' commands do this - your would use RSH from the VAX to run commands on a UNIX system, and use rsh on the UNIX system to run commands on the VAX. These are pretty standard features for VMS TCP/IP packages, and they come with UNIX TCP/IP as well. -- Steve Huston CompuCraft CompuServe: 72133,2372 61 Nadeau Drive Internet: 72133.2372@CompuServe.COM Wrentham, MA 02093-1618 huston@icecub.ako.dec.com (508) 384-5770 ================================================================================ Archive-Date: XXX, 5 Jul 1994 12:00:45 GMT Subject: tcp-ware performance between VAX-UNIX Message-ID: <2vbi1d$n83@pheidippides.axion.bt.co.uk> From: phendy@axion.bt.co.uk (Paul Hendy) Date: 5 Jul 1994 12:00:45 GMT Keywords: performance Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Does anyone have any experience of: 1) performance overheads associated with using tcp-ware between VAX-UNIX 2) monitoring tcp activity 3) are there any tunable parameters, buffer sizes, global areas etc. We are using tcp-ware to link a VAX based application to obtain sqlnet access to ORACLE on a UNIX (Sequent server). We find that the CPU cost per transaction of the application increases as the overall workload increases. i.e. the same transaction is costing more CPU as the system gets busier. We have used PCA (to scan the running code) and most of the CPU time is being burnt in sqlnet and lower level to link to the server. The buffered I/O is also increasing as well as CPU. Does tcp-ware keep a limited buffer space? so when more requests are active it has to do many more buffered I/Os and hence more CPU. Also, is the tcp-ware link interrupt driven, or is there an overhead of polling from the UNIX end. This problem means our systems performance is not scalable, i.e. the busier it gets even more processing power will be needed. Any ideas please. -- ********************************************************* * Paul Hendy * * Management Systems Performance & Development * * MLB 5/51J * * BT Labs, Martlesham Heath * * * * Tel. 0473 - 645818 * * Fax. 0473 - 640279 * ********************************************************* ================================================================================ Archive-Date: XXX, 6 Jul 1994 22:45:34 -0400 From: volz@process.com (Bernie Volz) Subject: Re: tcp-ware performance between VAX-UNIX Message-ID: <1994Jul6.224534.557@process.com> Date: 6 Jul 94 22:45:34 -0400 References: <2vbi1d$n83@pheidippides.axion.bt.co.uk> In article <2vbi1d$n83@pheidippides.axion.bt.co.uk>, phendy@axion.bt.co.uk (Paul Hendy) writes: > Does anyone have any experience of: > > 1) performance overheads associated with using tcp-ware between > VAX-UNIX > 2) monitoring tcp activity > 3) are there any tunable parameters, buffer sizes, global areas etc. > ... > > The buffered I/O is also increasing as well as CPU. Does tcp-ware > keep a limited buffer space? so when more requests are active it has > to do many more buffered I/Os and hence more CPU. All network I/O over TCPware is buffered-I/O. So, the more activity you have, the more buffered-I/O will occur. We do attempt to keep buffer copies to a minimum - the receive code is very well optimized and usually only entails copying the data into the user's buffer (and there is never any way around that). While the load should scale fairly well, there are some additional costs associated with more connections. One, for example, is locating the data structure for a connection costs more as the number of connections increase. Another is timer based scans - more connections means more time spent going through the data structures to see if anything timed out. TCPware makes direct use of VMS non-paged pool (it does not keep its own pool). This means that tuning pool is probably the best thing you can do. Make sure your IRP, SRP, LRP, and non-paged pool parameters are set to reasonable values; run it for a while and AUTOGEN w/feedback. If you expect the load to increase, increase the values by 25% or so. Depending on how SQLNET works, perhaps it creates more processes to handle more load. This also means more context switches and general overhead. > > Also, is the tcp-ware link interrupt driven, or is there an overhead > of polling from the UNIX end. Yes, TCPware is completely interrupt driven. No polling at all. It is "interrupted" when packets are received and when timers expire. > > This problem means our systems performance is not scalable, i.e. the > busier it gets even more processing power will be needed. > > Any ideas please. > Another recommendation is to make certain you are using the latest version of TCPware. We do work on every release to improve the performance. TCPware for OpenVMS V4.1 is just coming out and it has some significant performance improvements (especially on AXP). - Bernie Volz Process Software Corporation ================================================================================ Archive-Date: XXX, 7 Jul 1994 11:27:02 +0100 Subject: printing over tcpware, problem with symbiont Message-ID: <1994Jul7.112702.50@gw2.admin.ch> From: schneider_h@gw2.admin.ch Date: 7 Jul 94 11:27:02 +0100 normally we works with dcps (lat) under vax-vms v5.5. now we have installed tcpware v4.0 and we would like to print over tcp/ip we used declaser-2200 printer when i initialize the new queue with the specification init/que/proc=tcpware_tssym/on="ip-address,port-address" printer nothing is doing i guess somthing is wrong with the symbiont has anyone the same problem or has anyone a solution Thanks Schneider Hans-Rudolf Section IDZ Bundesamt für Informatik CH-3003 Bern ================================================================================ Archive-Date: XXX, 7 Jul 1994 10:34:43 -0400 From: volz@process.com (Bernie Volz) Subject: Re: printing over tcpware, problem with symbiont Message-ID: <1994Jul7.103443.558@process.com> Date: 7 Jul 94 10:34:43 -0400 References: <1994Jul7.112702.50@gw2.admin.ch> In article <1994Jul7.112702.50@gw2.admin.ch>, schneider_h@gw2.admin.ch writes: > normally we works with dcps (lat) under vax-vms v5.5. > now we have installed tcpware v4.0 and we would like to print over tcp/ip > we used declaser-2200 printer > when i initialize the new queue with the specification > init/que/proc=tcpware_tssym/on="ip-address,port-address" printer > nothing is doing > i guess somthing is wrong with the symbiont > has anyone the same problem or has anyone a solution > Could you be a bit more specific about what is happening? And, I assume you do a START/QUEUE (or INIT/QUEUE/START)? It the queue is started and it can't connect, it will send messages to OPCOM. It would probably be best to open a support call - send email to support@process.com or call our support group at 508-879-6994. - Bernie Volz Process Software Corporation ================================================================================ Archive-Date: Wed, 13 Jul 1994 15:18:07 GMT Subject: HELP! Need to change hostnames Message-ID: <1994Jul13.151807.19063@njitgw.njit.edu> From: rxy5310@hertz.njit.edu (Rolan) Date: Wed, 13 Jul 1994 15:18:07 GMT Sender: news@njit.edu Hello, I am in need of some help. I am running a TCP/IP network with PC-NFS software. There are about 15 PC's linked to a SUN. I am trying to change the hostnames on each computer. I changed the name using the DIRECT CONNECT option in nfsconf.exe to no avail. I also edited the NETWORK.BAT file. Still no change. I also changed the /etc/hosts file. No luck. Please drop me a line if anyone has suggestions. -- ------------------------------------------------------------------------------- Rolan Yang rxy5310@hertz.njit.edu Electrical Engineering New Jersey Institute of Technology -------------------------------------------------------------------------------