Hyper threading
A DBA moving Informix (12.10 and 9.40 on AIX 6/7) from Power7 to Power8 asked whether SMT/hyperthreading should be enabled and how to tune for it. No formal documentation was offered; the advice was experiential: Paul Watson found throughput roughly the same with SMT on or off (the server just loads up sooner with it off) and urged testing since SMT is easy to change. Art Kagel recommended ignoring SMT threads and sizing by physical cores, disabling or reducing SMT on P6/P7, and allowing ~2 CPU VPs per GHz (more VPs than cores), caveated by I/O bandwidth, user counts and DW parallelism. Fernando Nunes warned that topas and nmon report CPU usage differently on AIX. No definitive P8 conclusion was reached; the poster noted his sysadmins control the setting anyway.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
IDS 12.10.fc5 IDS 9.40.fc5 Aix 6 and 7 We have just moved our Informix systems to a Power 8 system ... previously we were on a Power 7. I remember something from last years conference about Hyper threading ... and if it should be used or not ... but I don't remember .. nor can I find it ... Does anyone have any guidelines as to the best performance ... Currently the system has 8 hyper threads per CPU. Also, any supporting documentation either way. Thanks for the help ... Peter Logan Database Administrator SpartanNash INC. Office: 616/878-8309 Mobile: 616/304-9672 E-Mail: peter.logan@spartannash.com
In my testing I see you can drive the server harder sooner with hyperthreading off but the absolute throughput is the same Cheers Paul Paul Watson Oninit www.oninit.com +1 913 387 7529 Oninit® is a Registered Trademark of Oninit LLC > On Oct 12, 2015, at 13:50, "Peter.Logan@spartannash.com" <Peter.Logan@spartannash.com> wrote: > > IDS 12.10.fc5 > IDS 9.40.fc5 > Aix 6 and 7 > We have just moved our Informix systems to a Power 8 system ... previously > we were on a Power 7. I remember something from last years conference > about Hyper threading ... and if it should be used or not ... but I don't > remember .. nor can I find it ... > > Does anyone have any guidelines as to the best performance ... Currently > the system has 8 hyper threads per CPU. Also, any supporting > documentation either way. > > Thanks for the help ... > > Peter Logan > Database Administrator > SpartanNash INC. > Office: 616/878-8309 > Mobile: 616/304-9672 > E-Mail: peter.logan@spartannash.com > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum.
Peter: I have not tested the P8s much, but on the P6 & P7s I have always found it best to ignore the SMT threads and deal with the actual number of physical cores only. The usual rule of thumb for the number of CPU VPs per core still apply. You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you should be able to run 6-8 CPU VPs per core with positive performance gains from the 6th, 7th, & 8th VP. As to having hte SMT threads configured and running, on P6 I have always recommended that clients disable them. On P7 which have 4 threads I recommend either running only two SMT threads per core (SMT=SMT2), disabling them altogether (SMT=ST) or even running the core in single threaded express-mode. I cannot say whether I would change these recommendations for the P8 chips as I haven't had any clients running P8 who were stressing their machines (and only one P8 user altogether). Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Mon, Oct 12, 2015 at 2:50 PM, Peter.Logan@spartannash.com < Peter.Logan@spartannash.com> wrote: > IDS 12.10.fc5 > IDS 9.40.fc5 > Aix 6 and 7 > We have just moved our Informix systems to a Power 8 system ... previously > we were on a Power 7. I remember something from last years conference > about Hyper threading ... and if it should be used or not ... but I don't > remember .. nor can I find it ... > > Does anyone have any guidelines as to the best performance ... Currently > the system has 8 hyper threads per CPU. Also, any supporting > documentation either way. > > Thanks for the help ... > > Peter Logan > Database Administrator > SpartanNash INC. > Office: 616/878-8309 > Mobile: 616/304-9672 > E-Mail: peter.logan@spartannash.com > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a114158b0fb70630521ed3fea
The nice thing with SMT is it trivial to change, so change it and test Cheers Paul Paul Watson Oninit www.oninit.com +1 913 387 7529 Oninit® is a Registered Trademark of Oninit LLC > On Oct 12, 2015, at 14:22, Art Kagel <art.kagel@gmail.com> wrote: > > Peter: > > I have not tested the P8s much, but on the P6 & P7s I have always found it > best to ignore the SMT threads and deal with the actual number of physical > cores only. > > The usual rule of thumb for the number of CPU VPs per core still apply. > You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you > should be able to run 6-8 CPU VPs per core with positive performance gains > from the 6th, 7th, & 8th VP. > > As to having hte SMT threads configured and running, on P6 I have always > recommended that clients disable them. On P7 which have 4 threads I > recommend either running only two SMT threads per core (SMT=SMT2), > disabling them altogether (SMT=ST) or even running the core in single > threaded express-mode. > > I cannot say whether I would change these recommendations for the P8 chips > as I haven't had any clients running P8 who were stressing their machines > (and only one P8 user altogether). > > Art > > Art S. Kagel, President and Principal Consultant > ASK Database Management > www.askdbmgt.com > > Blog: http://informix-myview.blogspot.com/ > > Disclaimer: Please keep in mind that my own opinions are my own opinions > and do not reflect on the IIUG, nor any other organization with which I am > associated either explicitly, implicitly, or by inference. Neither do > those opinions reflect those of other individuals affiliated with any > entity with which I am affiliated nor those of the entities themselves. > > On Mon, Oct 12, 2015 at 2:50 PM, Peter.Logan@spartannash.com < > Peter.Logan@spartannash.com> wrote: > >> IDS 12.10.fc5 >> IDS 9.40.fc5 >> Aix 6 and 7 >> We have just moved our Informix systems to a Power 8 system ... previously >> we were on a Power 7. I remember something from last years conference >> about Hyper threading ... and if it should be used or not ... but I don't >> remember .. nor can I find it ... >> >> Does anyone have any guidelines as to the best performance ... Currently >> the system has 8 hyper threads per CPU. Also, any supporting >> documentation either way. >> >> Thanks for the help ... >> >> Peter Logan >> Database Administrator >> SpartanNash INC. >> Office: 616/878-8309 >> Mobile: 616/304-9672 >> E-Mail: peter.logan@spartannash.com > ******************************************************************************* >> Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a114158b0fb70630521ed3fea > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum.
Art I'm interested in your comments "You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you should be able to run 6-8 CPU VPs per core with positive performance gains from the 6th, 7th, & 8th VP. " My simple understanding, based on documentation is "On systems, where multi-core processors are installed, the optimal configuration in most cases is the same as for systems with a number of individual processors equal to the total number of cores. Setting the number of CPU VPs to N-1, where N is number of cores is close to optimal for CPU-intensive workloads." So if I understand your statement correctly we could have more CPU VP's than cores available. So if we're looking at a P8 with two 12-core 3.52 GHz POWER8 processor cards we could in theory run 6 - 7 CPU VP's per core or 24 times this number. Of course we would not do this as the O/S needs some capacity. Thanks in advance. Jeff Poulton
Jeff: Yes, I disagree with IBM based on testing I and several other DBAs did in the mid '90s and my experiences since. I find that a CPU VP can't take advantage of much more than about 500MHZ of processor cycles, so if you want to get maximum throughput and responsiveness out of a systems for a large number of active users, then configure two CPU VPs per GHZ. Example: I was just looking at a client's machine. They have 8x2.4GHZ processor cores and were running 8 CPU VPs. I ran top and the 8 CPU VPs were running at the top of the process list constantly each using about 20% of a CPU core and the machine was 88% idle. I added 8 more CPU VPs, idle dropped only to between 80 & 84% but now the top 6 CPU VPs were only using about 15% of a core with the bottom 6 using about 9%. More work was getting done concurrently, users are experiencing a system that seems to be more responsive and so faster, and nothing was lost. Another client, a couple of months ago. They were running 5 cores with 20 CPU VPs (up from 5 before I got there) and were getting throughput from 150 sessions ranked at ~100 complex transactions/sec. They added another 5 cores but initially didn't add any CPU VPs, so 10 cores and 20 VPs. Throughput rose a moderate 20% to 120 transactions/sec. I added another 20 CPU VPs so 20 cores and 40 VPs and throughput increased to over 200 transactions/sec. QED Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Mon, Oct 12, 2015 at 8:28 PM, <> wrote: > Art > > I'm interested in your comments > > "You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you > should be able to run 6-8 CPU VPs per core with positive performance gains > from the 6th, 7th, & 8th VP. " > > My simple understanding, based on documentation is > > "On systems, where multi-core processors are installed, the optimal > configuration in most cases is the same as for systems with a number of > individual processors equal to the total number of cores. Setting the > number > of CPU VPs to N-1, where N is number of cores is close to optimal for > CPU-intensive workloads." > > So if I understand your statement correctly we could have more CPU VP's > than > cores available. So if we're looking at a P8 with two 12-core 3.52 GHz > POWER8 > processor cards we could in theory run 6 - 7 CPU VP's per core or 24 times > this number. Of course we would not do this as the O/S needs some capacity. > > Thanks in advance. > > Jeff Poulton > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a11349cee966bb50521f1e015
Oops that should be 10 cores and 40 VPs in the final configuration. Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Mon, Oct 12, 2015 at 8:53 PM, Art Kagel <art.kagel@gmail.com> wrote: > Jeff: > > Yes, I disagree with IBM based on testing I and several other DBAs did in > the mid '90s and my experiences since. I find that a CPU VP can't take > advantage of much more than about 500MHZ of processor cycles, so if you > want to get maximum throughput and responsiveness out of a systems for a > large number of active users, then configure two CPU VPs per GHZ. > > Example: I was just looking at a client's machine. They have 8x2.4GHZ > processor cores and were running 8 CPU VPs. I ran top and the 8 CPU VPs > were running at the top of the process list constantly each using about 20% > of a CPU core and the machine was 88% idle. I added 8 more CPU VPs, idle > dropped only to between 80 & 84% but now the top 6 CPU VPs were only using > about 15% of a core with the bottom 6 using about 9%. More work was > getting done concurrently, users are experiencing a system that seems to be > more responsive and so faster, and nothing was lost. > > Another client, a couple of months ago. They were running 5 cores with 20 > CPU VPs (up from 5 before I got there) and were getting throughput from 150 > sessions ranked at ~100 complex transactions/sec. They added another 5 > cores but initially didn't add any CPU VPs, so 10 cores and 20 VPs. > Throughput rose a moderate 20% to 120 transactions/sec. I added another 20 > CPU VPs so 20 cores and 40 VPs and throughput increased to over 200 > transactions/sec. QED > > Art > > > Art S. Kagel, President and Principal Consultant > ASK Database Management > www.askdbmgt.com > > Blog: http://informix-myview.blogspot.com/ > > Disclaimer: Please keep in mind that my own opinions are my own opinions > and do not reflect on the IIUG, nor any other organization with which I am > associated either explicitly, implicitly, or by inference. Neither do > those opinions reflect those of other individuals affiliated with any > entity with which I am affiliated nor those of the entities themselves. > > > On Mon, Oct 12, 2015 at 8:28 PM, <> wrote: > >> Art >> >> I'm interested in your comments >> >> "You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor >> you >> should be able to run 6-8 CPU VPs per core with positive performance gains >> from the 6th, 7th, & 8th VP. " >> >> My simple understanding, based on documentation is >> >> "On systems, where multi-core processors are installed, the optimal >> configuration in most cases is the same as for systems with a number of >> individual processors equal to the total number of cores. Setting the >> number >> of CPU VPs to N-1, where N is number of cores is close to optimal for >> CPU-intensive workloads." >> >> So if I understand your statement correctly we could have more CPU VP's >> than >> cores available. So if we're looking at a P8 with two 12-core 3.52 GHz >> POWER8 >> processor cards we could in theory run 6 - 7 CPU VP's per core or 24 times >> this number. Of course we would not do this as the O/S needs some >> capacity. >> >> Thanks in advance. >> >> Jeff Poulton >> >> >> >> ******************************************************************************* >> Forum Note: Use "Reply" to post a response in the discussion forum. >> >> > --089e01229aaa916cb60521f1e4c0
Art Thanks for the reply. Some food for thought. Regards Jeff
You're welcome. Someone reminded me privately that one thing that can affect the scalability of the CPU VPs is IO bandwidth. If the chunks cannot keep up with the added transaction submission from more CPU VPs, then performance will not increase with more VPs and may actually decrease somewhat. Similarly, if you only have a few users then there is limited benefit to having many CPU VPs. Finally, data warehouse systems should be tuned to the amount of parallelism that the schema structure (ie table & index partitioning) can take advantage of rather than pure CPU cycles. Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Mon, Oct 12, 2015 at 10:28 PM, <> wrote: > Art > > Thanks for the reply. Some food for thought. > > Regards > > Jeff > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a114158b0f0d7360521f3487b
Art .. Thanks for the info ... Sadly our Unix guys do there own thing ... so, I get to stay with what they set ... and they of course are the experts ... ha ... Peter Logan Database Administrator SpartanNash INC. Office: 616/878-8309 Mobile: 616/304-9672 E-Mail: peter.logan@spartannash.com From: "Art Kagel" <art.kagel@gmail.com> To: ids@iiug.org Date: 10/12/2015 03:22 PM Subject: Re: Hyper threading [35858] Sent by: ids-bounces@iiug.org Peter: I have not tested the P8s much, but on the P6 & P7s I have always found it best to ignore the SMT threads and deal with the actual number of physical cores only. The usual rule of thumb for the number of CPU VPs per core still apply. You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you should be able to run 6-8 CPU VPs per core with positive performance gains from the 6th, 7th, & 8th VP. As to having hte SMT threads configured and running, on P6 I have always recommended that clients disable them. On P7 which have 4 threads I recommend either running only two SMT threads per core (SMT=SMT2), disabling them altogether (SMT=ST) or even running the core in single threaded express-mode. I cannot say whether I would change these recommendations for the P8 chips as I haven't had any clients running P8 who were stressing their machines (and only one P8 user altogether). Art Art S. Kagel, President and Principal Consultant ASK Database Management www.askdbmgt.com Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Mon, Oct 12, 2015 at 2:50 PM, Peter.Logan@spartannash.com < Peter.Logan@spartannash.com> wrote: > IDS 12.10.fc5 > IDS 9.40.fc5 > Aix 6 and 7 > We have just moved our Informix systems to a Power 8 system ... previously > we were on a Power 7. I remember something from last years conference > about Hyper threading ... and if it should be used or not ... but I don't > remember .. nor can I find it ... > > Does anyone have any guidelines as to the best performance ... Currently > the system has 8 hyper threads per CPU. Also, any supporting > documentation either way. > > Thanks for the help ... > > Peter Logan > Database Administrator > SpartanNash INC. > Office: 616/878-8309 > Mobile: 616/304-9672 > E-Mail: peter.logan@spartannash.com > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a114158b0fb70630521ed3fea ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.
If you root then you always test independently :-). However, I know of two sites where having tested and shown the DBA setup was faster they still couldn't get the changes implemented. Cheers Paul Paul Watson Oninit www.oninit.com +1 913 387 7529 Oninit® is a Registered Trademark of Oninit LLC > On Oct 13, 2015, at 06:38, "Peter.Logan@spartannash.com" <Peter.Logan@spartannash.com> wrote: > > Art .. > > Thanks for the info ... Sadly our Unix guys do there own thing ... so, I > get to stay with what they set ... and they of course are the experts ... > ha ... > > Peter Logan > Database Administrator > SpartanNash INC. > Office: 616/878-8309 > Mobile: 616/304-9672 > E-Mail: peter.logan@spartannash.com > > From: "Art Kagel" <art.kagel@gmail.com> > To: ids@iiug.org > Date: 10/12/2015 03:22 PM > Subject: Re: Hyper threading [35858] > Sent by: ids-bounces@iiug.org > > Peter: > > I have not tested the P8s much, but on the P6 & P7s I have always found it > > best to ignore the SMT threads and deal with the actual number of physical > > cores only. > > The usual rule of thumb for the number of CPU VPs per core still apply. > You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor you > > should be able to run 6-8 CPU VPs per core with positive performance gains > > from the 6th, 7th, & 8th VP. > > As to having hte SMT threads configured and running, on P6 I have always > recommended that clients disable them. On P7 which have 4 threads I > recommend either running only two SMT threads per core (SMT=SMT2), > disabling them altogether (SMT=ST) or even running the core in single > threaded express-mode. > > I cannot say whether I would change these recommendations for the P8 chips > > as I haven't had any clients running P8 who were stressing their machines > (and only one P8 user altogether). > > Art > > Art S. Kagel, President and Principal Consultant > ASK Database Management > www.askdbmgt.com > > Blog: http://informix-myview.blogspot.com/ > > Disclaimer: Please keep in mind that my own opinions are my own opinions > and do not reflect on the IIUG, nor any other organization with which I am > > associated either explicitly, implicitly, or by inference. Neither do > those opinions reflect those of other individuals affiliated with any > entity with which I am affiliated nor those of the entities themselves. > > On Mon, Oct 12, 2015 at 2:50 PM, Peter.Logan@spartannash.com < > Peter.Logan@spartannash.com> wrote: > >> IDS 12.10.fc5 >> IDS 9.40.fc5 >> Aix 6 and 7 >> We have just moved our Informix systems to a Power 8 system ... > previously >> we were on a Power 7. I remember something from last years conference >> about Hyper threading ... and if it should be used or not ... but I > don't >> remember .. nor can I find it ... >> >> Does anyone have any guidelines as to the best performance ... Currently > >> the system has 8 hyper threads per CPU. Also, any supporting >> documentation either way. >> >> Thanks for the help ... >> >> Peter Logan >> Database Administrator >> SpartanNash INC. >> Office: 616/878-8309 >> Mobile: 616/304-9672 >> E-Mail: peter.logan@spartannash.com > > ******************************************************************************* > >> Forum Note: Use "Reply" to post a response in the discussion forum. > > --001a114158b0fb70630521ed3fea > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum.
One note for AIX. Not sure if the case Art mention was AIX, but the OP mentioned AIX. You'll notice a big difference between data showed by topas and nmon. I will not get into details, mainly because i think there are better sources of information on the Internet, but from what I recall, topas shows the cpu usage percentage compared to the entitlement of the LPAR. nmon shows the percentage of time the process was active on the CPU. This can make a huge difference and cause a lot of misinterpretation of values. On a recent situation this was between 7% and 50-60%. So at first the suggestion was "we have I/O issues, because the process is not consuming enough CPU". After seeing the values from nmon this changed to: We've reached the maximum throughput. We need faster I/O or probably we need to parallelize the process (this was not related to Informix) So, be careful. Also, in my experience the sysadmins are not well aware of these differences. And they tend to create explanations for whatever is suggested (if there's any AIX sysadmin over there, fell free to bash me, but I'm reporting my own experience and this is not specific for AIX) Regards. On Tue, Oct 13, 2015 at 1:53 AM, Art Kagel <art.kagel@gmail.com> wrote: > Jeff: > > Yes, I disagree with IBM based on testing I and several other DBAs did in > the mid '90s and my experiences since. I find that a CPU VP can't take > advantage of much more than about 500MHZ of processor cycles, so if you > want to get maximum throughput and responsiveness out of a systems for a > large number of active users, then configure two CPU VPs per GHZ. > > Example: I was just looking at a client's machine. They have 8x2.4GHZ > processor cores and were running 8 CPU VPs. I ran top and the 8 CPU VPs > were running at the top of the process list constantly each using about 20% > of a CPU core and the machine was 88% idle. I added 8 more CPU VPs, idle > dropped only to between 80 & 84% but now the top 6 CPU VPs were only using > about 15% of a core with the bottom 6 using about 9%. More work was > getting done concurrently, users are experiencing a system that seems to be > more responsive and so faster, and nothing was lost. > > Another client, a couple of months ago. They were running 5 cores with 20 > CPU VPs (up from 5 before I got there) and were getting throughput from 150 > sessions ranked at ~100 complex transactions/sec. They added another 5 > cores but initially didn't add any CPU VPs, so 10 cores and 20 VPs. > Throughput rose a moderate 20% to 120 transactions/sec. I added another 20 > CPU VPs so 20 cores and 40 VPs and throughput increased to over 200 > transactions/sec. QED > > Art > > Art S. Kagel, President and Principal Consultant > ASK Database Management > www.askdbmgt.com > > Blog: http://informix-myview.blogspot.com/ > > Disclaimer: Please keep in mind that my own opinions are my own opinions > and do not reflect on the IIUG, nor any other organization with which I am > associated either explicitly, implicitly, or by inference. Neither do > those opinions reflect those of other individuals affiliated with any > entity with which I am affiliated nor those of the entities themselves. > > On Mon, Oct 12, 2015 at 8:28 PM, <> wrote: > > > Art > > > > I'm interested in your comments > > > > "You need about 500MHZ of cpu cycles per CPU VP so for a 4GHZ processor > you > > should be able to run 6-8 CPU VPs per core with positive performance > gains > > from the 6th, 7th, & 8th VP. " > > > > My simple understanding, based on documentation is > > > > "On systems, where multi-core processors are installed, the optimal > > configuration in most cases is the same as for systems with a number of > > individual processors equal to the total number of cores. Setting the > > number > > of CPU VPs to N-1, where N is number of cores is close to optimal for > > CPU-intensive workloads." > > > > So if I understand your statement correctly we could have more CPU VP's > > than > > cores available. So if we're looking at a P8 with two 12-core 3.52 GHz > > POWER8 > > processor cards we could in theory run 6 - 7 CPU VP's per core or 24 > times > > this number. Of course we would not do this as the O/S needs some > capacity. > > > > Thanks in advance. > > > > Jeff Poulton > > > > > > > > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > --001a11349cee966bb50521f1e015 > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > -- Fernando Nunes Portugal http://informix-technology.blogspot.com My email works... but I don't check it frequently... --089e013a01c8bb12350521fe268e