Re: A Question on Sizing z/VM Linux Guests
Another handy item for Oracle folks is: http://www.oracle.com/technology/software/tech/orion/index.html Allows you to simulate Oracle I/O loads on disk configurations without actually having to install Oracle and find a database big enough to stress test your filesystem configuration and get your DBAs involved (at least at first). It also gives you an apples-to-apples comparison with different implementations on different I/O architectures. Very helpful in finding Oracle disk performance problems. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
Thanks for the comments. > to set DB2 memory management to "automatic" to let it size itself based on the Linux configuration. Which can be interpreted as, with STMM set (fully) on, DB2 will try to use 85% of configured memory (if it needs it ofr not). Then it's back to guessing how much is really needed. After the initial wag of sizing, (hopefully it is on the conservative side), it finding the point where there is enough memory to run at the peak loads and have a little to spare and not too much to cause excessive VM paging. I am a little surprised that we are not mentioning response times for DB requests during peak periods, or throughput as metrics to help evaluate memory needs and the tools/ways to help collect this data. Thx Rob van der Heij Sent by: Linux on 390 Port 02/05/2010 10:03 AM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Re: A Question on Sizing z/VM Linux Guests On Fri, Feb 5, 2010 at 4:26 PM, Sterling James wrote: > I realize that it would not be a apples to apples comparison, but is there > a similar discussion about sizing guests that are running DB2? With STMM > on? And the "Be ware!"s? Does DB2 have instrumentation that can tell you > how it's using memory and indicate "good" value. I am aware the STMM is > suppose to do that automatically, but I am suspicious of it's assumption > since it does know account for running under zVM. The DB2 folks will prefer to compare oranges and oranges, since IBM is very strict on not involving competitors in comparisons. :-) You're right, with DB2 there's also shared memory between threads that lives in page cache. So you should expect the page cache to be fairly large on such a system. I did find that many DBA's seem to set DB2 memory management to "automatic" to let it size itself based on the Linux configuration. I don't think DB2 needs to be aware that Linux runs on z/VM. If the automatic memory management works well, it would not be a bad approach that you'd only have to size the virtual machine to make it happen. We recently had conversations with the DB2 people discussing changes in DB2 to improve behavior for low-utilized servers with Linux on z/VM. I'm supposed to get my hands on the fix pack that implements those RSN and be able to kick the tires. I'm also interested to investigate the tuning options and instrumentation. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 - Please consider the environment before printing this email and any attachments. This e-mail and any attachments are intended only for the individual or company to which it is addressed and may contain information which is privileged, confidential and prohibited from disclosure or unauthorized use under applicable law. If you are not the intended recipient of this e-mail, you are hereby notified that any use, dissemination, or copying of this e-mail or the information contained in this e-mail is strictly prohibited by the sender. If you have received this transmission in error, please return the material received to the sender and delete all copies from your system. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
On Fri, Feb 5, 2010 at 4:26 PM, Sterling James wrote: > I realize that it would not be a apples to apples comparison, but is there > a similar discussion about sizing guests that are running DB2? With STMM > on? And the "Be ware!"s? Does DB2 have instrumentation that can tell you > how it's using memory and indicate "good" value. I am aware the STMM is > suppose to do that automatically, but I am suspicious of it's assumption > since it does know account for running under zVM. The DB2 folks will prefer to compare oranges and oranges, since IBM is very strict on not involving competitors in comparisons. :-) You're right, with DB2 there's also shared memory between threads that lives in page cache. So you should expect the page cache to be fairly large on such a system. I did find that many DBA's seem to set DB2 memory management to "automatic" to let it size itself based on the Linux configuration. I don't think DB2 needs to be aware that Linux runs on z/VM. If the automatic memory management works well, it would not be a bad approach that you'd only have to size the virtual machine to make it happen. We recently had conversations with the DB2 people discussing changes in DB2 to improve behavior for low-utilized servers with Linux on z/VM. I'm supposed to get my hands on the fix pack that implements those RSN and be able to kick the tires. I'm also interested to investigate the tuning options and instrumentation. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
I realize that it would not be a apples to apples comparison, but is there a similar discussion about sizing guests that are running DB2? With STMM on? And the "Be ware!"s? Does DB2 have instrumentation that can tell you how it's using memory and indicate "good" value. I am aware the STMM is suppose to do that automatically, but I am suspicious of it's assumption since it does know account for running under zVM. Thx Rob van der Heij Sent by: Linux on 390 Port 02/05/2010 06:14 AM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Re: A Question on Sizing z/VM Linux Guests 2010/2/4 van Sleeuwen, Berry : > We have some oracle guests that have been set to 1G. And they actually > need only 300M. The rest is occupied in cache (between 650 and 710M). Be aware! The Oracle SGA lives in Linux page cache, together with in-use programs and shared libraries. So on Oracle systems it is normal to see a large page cache and it is not all waste. The Oracle instrumentation can tell you how much of the allowed SGA size is actually being use (and whether that is enough). When the system sees enough data, it will eventually grow to use all allowed SGA space. Without access to those metrics, you could write a "1" into the /proc/sys/vm/drop_caches to have Linux drop all other stuff from page cache. You will see Linux reclaim some of that in a few minutes, and that might be a good stake in the ground. > I'd like to decrease the machines but it is quite hard to convince the > oracle group that the machines do not need the memory. Instead, they > even want to increase to 2G because the books tell them to. While the > documents suggest the oracle will benefit from the increased memory, in > reality VM will page more and the effect is an overall decrease in > performance for all guests. You're correct that there is a cost for increasing the virtual machine size, but there is also a cost for not increasing it. You need to measure and analyze that. And you have the product that that can help you understand the cost rather than tune by fear and doubt :-) There's always a trade-off and your analysis should include the utilization of the virtual machine. When the server is rarely active, it may be acceptable when it is less efficient since that reduces the resource cost when it is idle. But servers that are very busy will benefit from efficient operation and you're less concerned about the idle footprint. > I would suggest to keep the guest as small as possible. Also try to > limit SGA and such. Among the "such" is also the PGA as sized by the DBA. That does not live in page cache but in anonymous memory in Linux (the "used" part that is not cache). Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 - Please consider the environment before printing this email and any attachments. This e-mail and any attachments are intended only for the individual or company to which it is addressed and may contain information which is privileged, confidential and prohibited from disclosure or unauthorized use under applicable law. If you are not the intended recipient of this e-mail, you are hereby notified that any use, dissemination, or copying of this e-mail or the information contained in this e-mail is strictly prohibited by the sender. If you have received this transmission in error, please return the material received to the sender and delete all copies from your system. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
O, I didn't know that. I'd expected the reported linux cache to be filesystem cache and as such not directly related to oracle. Berry. -Original Message- Be aware! The Oracle SGA lives in Linux page cache, together with in-use programs and shared libraries. So on Oracle systems it is normal to see a large page cache and it is not all waste. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ÿþD i t b e r i c h t i s v e r t r o u w e l i j k e n k a n g e h e i m e i n f o r m a t i e b e v a t t e n e n k e l b e s t e m d v o o r d e g e a d r e s s e e r d e . I n d i e n d i t b e r i c h t n i e t v o o r u i s b e s t e m d , v e r z o e k e n w i j u d i t o n m i d d e l l i j k a a n o n s t e m e l d e n e n h e t b e r i c h t t e v e r n i e t i g e n . A a n g e z i e n d e i n t e g r i t e i t v a n h e t b e r i c h t n i e t v e i l i g g e s t e l d i s m i d d e l s v e r z e n d i n g v i a i n t e r n e t , k a n A t o s O r i g i n n i e t a a n s p r a k e l i j k w o r d e n g e h o u d e n v o o r d e i n h o u d d a a r v a n . H o e w e l w i j o n s i n s p a n n e n e e n v i r u s v r i j n e t w e r k t e h a n t e r e n , g e v e n w i j g e e n e n k e l e g a r a n t i e d a t d i t b e r i c h t v i r u s v r i j i s , n o c h a a n v a a r d e n w i j e n i g e a a n s p r a k e l i j k h e i d v o o r d e m o g e l i j k e a a n w e z i g h e i d v a n e e n v i r u s i n d i t b e r i c h t . O p a l o n z e r e c h t s v e r h o u d i n g e n , a a n b i e d i n g e n e n o v e r e e n k o m s t e n w a a r o n d e r A t o s O r i g i n g o e d e r e n e n / o f d i e n s t e n l e v e r t z i j n m e t u i t s l u i t i n g v a n a l l e a n d e r e v o o r w a a r d e n d e L e v e r i n g s v o o r w a a r d e n v a n A t o s O r i g i n v a n t o e p a s s i n g . D e z e w o r d e n u o p a a n v r a a g d i r e c t k o s t e l o o s t o e g e z o n d e n . T h i s e - m a i l a n d t h e d o c u m e n t s a t t a c h e d a r e c o n f i d e n t i a l a n d i n t e n d e d s o l e l y f o r t h e a d d r e s s e e ; i t m a y a l s o b e p r i v i l e g e d . I f y o u r e c e i v e t h i s e - m a i l i n e r r o r , p l e a s e n o t i f y t h e s e n d e r i m m e d i a t e l y a n d d e s t r o y i t . A s i t s i n t e g r i t y c a n n o t b e s e c u r e d o n t h e I n t e r n e t , t h e A t o s O r i g i n g r o u p l i a b i l i t y c a n n o t b e t r i g g e r e d f o r t h e m e s s a g e c o n t e n t . A l t h o u g h t h e s e n d e r e n d e a v o u r s t o m a i n t a i n a c o m p u t e r v i r u s - f r e e n e t w o r k , t h e s e n d e r d o e s n o t w a r r a n t t h a t t h i s t r a n s m i s s i o n i s v i r u s - f r e e a n d w i l l n o t b e l i a b l e f o r a n y d a m a g e s r e s u l t i n g f r o m a n y v i r u s t r a n s m i t t e d . O n a l l o f f e r s a n d a g r e e m e n t s u n d e r w h i c h A t o s O r i g i n s u p p l i e s g o o d s a n d / o r s e r v i c e s o f w h a t e v e r n a t u r e , t h e T e r m s o f D e l i v e r y f r o m A t o s O r i g i n e x c l u s i v e l y a p p l y . T h e T e r m s o f D e l i v e r y s h a l l b e p r o m p t l y s u b m i t t e d t o y o u o n y o u r r e q u e s t . A t o s O r i g i n N e d e r l a n d B . V . / U t r e c h t K v K U t r e c h t 3 0 1 3 2 7 6 2
Re: A Question on Sizing z/VM Linux Guests
2010/2/4 van Sleeuwen, Berry : > We have some oracle guests that have been set to 1G. And they actually > need only 300M. The rest is occupied in cache (between 650 and 710M). Be aware! The Oracle SGA lives in Linux page cache, together with in-use programs and shared libraries. So on Oracle systems it is normal to see a large page cache and it is not all waste. The Oracle instrumentation can tell you how much of the allowed SGA size is actually being use (and whether that is enough). When the system sees enough data, it will eventually grow to use all allowed SGA space. Without access to those metrics, you could write a "1" into the /proc/sys/vm/drop_caches to have Linux drop all other stuff from page cache. You will see Linux reclaim some of that in a few minutes, and that might be a good stake in the ground. > I'd like to decrease the machines but it is quite hard to convince the > oracle group that the machines do not need the memory. Instead, they > even want to increase to 2G because the books tell them to. While the > documents suggest the oracle will benefit from the increased memory, in > reality VM will page more and the effect is an overall decrease in > performance for all guests. You're correct that there is a cost for increasing the virtual machine size, but there is also a cost for not increasing it. You need to measure and analyze that. And you have the product that that can help you understand the cost rather than tune by fear and doubt :-) There's always a trade-off and your analysis should include the utilization of the virtual machine. When the server is rarely active, it may be acceptable when it is less efficient since that reduces the resource cost when it is idle. But servers that are very busy will benefit from efficient operation and you're less concerned about the idle footprint. > I would suggest to keep the guest as small as possible. Also try to > limit SGA and such. Among the "such" is also the PGA as sized by the DBA. That does not live in page cache but in anonymous memory in Linux (the "used" part that is not cache). Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
I run my Oracle guests GT 90% mem and use Velocity to watch the WSS size and +- SGA for good performance on s390x. This approach is opposite of what (as you may of heard) you may do to improve performance on I686-RHEL. NONE of the individual guest DBs are over 4.0G MEM. Here is a recent tuning change which droped IO in the guest over 100% for a 4G guest. SQL> alter system set SGA_TARGET=1200M scope both; SQL> alter system set PGA_TARGET=600M scope both; SQL> alter system set SGA_MAX_SIZE=1200M scope both; SQL> alter system set DB_CACHE_SIZE=500M scope both; ://Gerard -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of Carson, Brad Sent: Thursday, February 04, 2010 8:01 AM To: LINUX-390@vm.marist.edu Subject: A Question on Sizing z/VM Linux Guests We have a project that is being setup to run on RHEL under z/VM. The sizing parameters for the guests (Oracle DB, and WebLogic) are being sent to us based on intel platform sizing. How do some of you handle the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? Our environment: IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. Thanks for any insight. /Brad (please ignore the company inserted HIPAA disclaimer) - This e-mail and any attachments may contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If you are not the intended recipient, any use or disclosure of this information is STRICTLY PROHIBITED; you are requested to delete this e-mail and any attachments, notify the sender immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
Yeah, but in this case it's the wrong book. With Oracle dropping support for their products on z/OS, the only mainframe platform now supported is Linux running as a guest of z/VM (or in an LPAR, I suppose). I think we'll soon see more z/VM-Linux specific performance guidelines from them in the future. Also take a look at the International zSeries Oracle Special Interest Group (SIG) which is focused on running Oracle applications and databases on zseries hardware. It can be found here: http://www.zseriesoraclesig.org/ They have all of their conference presentations available online, including the ones from our very own Barton Robinson. On 02/04/2010 09:02 AM, Ron Wells wrote: sounds familiar>> the book tells them to.. From: "van Sleeuwen, Berry" To: LINUX-390@VM.MARIST.EDU Date: 02/04/2010 08:32 AM Subject: Re: A Question on Sizing z/VM Linux Guests Sent by: Linux on 390 Port We have some oracle guests that have been set to 1G. And they actually need only 300M. The rest is occupied in cache (between 650 and 710M). I'd like to decrease the machines but it is quite hard to convince the oracle group that the machines do not need the memory. Instead, they even want to increase to 2G because the books tell them to. While the documents suggest the oracle will benefit from the increased memory, in reality VM will page more and the effect is an overall decrease in performance for all guests. I would suggest to keep the guest as small as possible. Also try to limit SGA and such. Berry. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of RPN01 Sent: donderdag 4 februari 2010 14:17 To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: We have a project that is being setup to run on RHEL under z/VM. The sizing parameters for the guests (Oracle DB, and WebLogic) are being sent to us based on intel platform sizing. How do some of you handle the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? Our environment: IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. Thanks for any insight. /Brad (please ignore the company inserted HIPAA disclaimer) - This e-mail and any attachments may contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If you are not the intended recipient, any use or disclosure of this information is STRICTLY PROHIBITED; you are requested to delete this e-mail and any attachments, notify the sender immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructio
Re: A Question on Sizing z/VM Linux Guests
Yep - I'd be interested to know which books if they are Oracle publications. -Original Message- From: Ron Wells [mailto:rwe...@agfinance.com] Sent: 04 February 2010 15:02 To: LINUX-390@vm.marist.edu Subject: Re: A Question on Sizing z/VM Linux Guests sounds familiar >> the book tells them to.. From: "van Sleeuwen, Berry" To: LINUX-390@VM.MARIST.EDU Date: 02/04/2010 08:32 AM Subject: Re: A Question on Sizing z/VM Linux Guests Sent by: Linux on 390 Port We have some oracle guests that have been set to 1G. And they actually need only 300M. The rest is occupied in cache (between 650 and 710M). I'd like to decrease the machines but it is quite hard to convince the oracle group that the machines do not need the memory. Instead, they even want to increase to 2G because the books tell them to. While the documents suggest the oracle will benefit from the increased memory, in reality VM will page more and the effect is an overall decrease in performance for all guests. I would suggest to keep the guest as small as possible. Also try to limit SGA and such. Berry. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of RPN01 Sent: donderdag 4 februari 2010 14:17 To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: > We have a project that is being setup to run on RHEL under z/VM. The > sizing parameters for the guests (Oracle DB, and WebLogic) are being > sent to us based on intel platform sizing. How do some of you handle > the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? > > Our environment: > > IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. > > > Thanks for any insight. > > /Brad (please ignore the company inserted HIPAA disclaimer) > > - This e-mail and any > attachments may contain CONFIDENTIAL information, including PROTECTED > HEALTH INFORMATION. If you are not the intended recipient, any use or > disclosure of this information is STRICTLY PROHIBITED; you are > requested to delete this e-mail and any attachments, notify the sender > immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. > > -- > For LINUX-390 subscribe / signoff / archive access instructions, send > email to lists...@vm.marist.edu with the message: INFO LINUX-390 or > visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send
Re: A Question on Sizing z/VM Linux Guests
sounds familiar >> the book tells them to.. From: "van Sleeuwen, Berry" To: LINUX-390@VM.MARIST.EDU Date: 02/04/2010 08:32 AM Subject: Re: A Question on Sizing z/VM Linux Guests Sent by: Linux on 390 Port We have some oracle guests that have been set to 1G. And they actually need only 300M. The rest is occupied in cache (between 650 and 710M). I'd like to decrease the machines but it is quite hard to convince the oracle group that the machines do not need the memory. Instead, they even want to increase to 2G because the books tell them to. While the documents suggest the oracle will benefit from the increased memory, in reality VM will page more and the effect is an overall decrease in performance for all guests. I would suggest to keep the guest as small as possible. Also try to limit SGA and such. Berry. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of RPN01 Sent: donderdag 4 februari 2010 14:17 To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: > We have a project that is being setup to run on RHEL under z/VM. The > sizing parameters for the guests (Oracle DB, and WebLogic) are being > sent to us based on intel platform sizing. How do some of you handle > the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? > > Our environment: > > IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. > > > Thanks for any insight. > > /Brad (please ignore the company inserted HIPAA disclaimer) > > - This e-mail and any > attachments may contain CONFIDENTIAL information, including PROTECTED > HEALTH INFORMATION. If you are not the intended recipient, any use or > disclosure of this information is STRICTLY PROHIBITED; you are > requested to delete this e-mail and any attachments, notify the sender > immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. > > -- > For LINUX-390 subscribe / signoff / archive access instructions, send > email to lists...@vm.marist.edu with the message: INFO LINUX-390 or > visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 [attachment "disclaimer.txt" deleted by Ron Wells/AGFS/AGFin] -- Email Disclaimer This E-mail contains confidentia
Re: A Question on Sizing z/VM Linux Guests
Berry, I'd be interested in seeing a 60 minute statspack or AWR/ADDM from one of these instances at peak hour - we generally recommend direct and asynch IO, which bypasses the cache. The reports will tell you what Oracle thinks it needs based on throughput, which you can compare to what z/VM and Linux is using. Contact me off list in the first place, as attachments don't survive well. Cheers Damian -Original Message- From: van Sleeuwen, Berry [mailto:berry.vansleeu...@atosorigin.com] Sent: 04 February 2010 14:32 To: LINUX-390@vm.marist.edu Subject: Re: A Question on Sizing z/VM Linux Guests We have some oracle guests that have been set to 1G. And they actually need only 300M. The rest is occupied in cache (between 650 and 710M). I'd like to decrease the machines but it is quite hard to convince the oracle group that the machines do not need the memory. Instead, they even want to increase to 2G because the books tell them to. While the documents suggest the oracle will benefit from the increased memory, in reality VM will page more and the effect is an overall decrease in performance for all guests. I would suggest to keep the guest as small as possible. Also try to limit SGA and such. Berry. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of RPN01 Sent: donderdag 4 februari 2010 14:17 To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: > We have a project that is being setup to run on RHEL under z/VM. The > sizing parameters for the guests (Oracle DB, and WebLogic) are being > sent to us based on intel platform sizing. How do some of you handle > the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? > > Our environment: > > IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. > > > Thanks for any insight. > > /Brad (please ignore the company inserted HIPAA disclaimer) > > - This e-mail and any > attachments may contain CONFIDENTIAL information, including PROTECTED > HEALTH INFORMATION. If you are not the intended recipient, any use or > disclosure of this information is STRICTLY PROHIBITED; you are > requested to delete this e-mail and any attachments, notify the sender > immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. > > -- > For LINUX-390 subscribe / signoff / archive access instructions, send > email to lists...@vm.marist.edu with the message: INFO LINUX-390 or > visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 --
Re: A Question on Sizing z/VM Linux Guests
We have some oracle guests that have been set to 1G. And they actually need only 300M. The rest is occupied in cache (between 650 and 710M). I'd like to decrease the machines but it is quite hard to convince the oracle group that the machines do not need the memory. Instead, they even want to increase to 2G because the books tell them to. While the documents suggest the oracle will benefit from the increased memory, in reality VM will page more and the effect is an overall decrease in performance for all guests. I would suggest to keep the guest as small as possible. Also try to limit SGA and such. Berry. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of RPN01 Sent: donderdag 4 februari 2010 14:17 To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: > We have a project that is being setup to run on RHEL under z/VM. The > sizing parameters for the guests (Oracle DB, and WebLogic) are being > sent to us based on intel platform sizing. How do some of you handle > the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? > > Our environment: > > IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. > > > Thanks for any insight. > > /Brad (please ignore the company inserted HIPAA disclaimer) > > - This e-mail and any > attachments may contain CONFIDENTIAL information, including PROTECTED > HEALTH INFORMATION. If you are not the intended recipient, any use or > disclosure of this information is STRICTLY PROHIBITED; you are > requested to delete this e-mail and any attachments, notify the sender > immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. > > -- > For LINUX-390 subscribe / signoff / archive access instructions, send > email to lists...@vm.marist.edu with the message: INFO LINUX-390 or > visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ÿþD i t b e r i c h t i s v e r t r o u w e l i j k e n k a n g e h e i m e i n f o r m a t i e b e v a t t e n e n k e l b e s t e m d v o o r d e g e a d r e s s e e r d e . I n d i e n d i t b e r i c h t n i e t v o o r u i s b e s t e m d , v e r z o e k e n w i j u d i t o n m i d d e l l i j k a a n o n s t e m e l d e n e n h e t b
Re: A Question on Sizing z/VM Linux Guests
If you're using IBM style DASD as your disk storage, you'll want to size the linux image's memory considerably smaller than what is suggested for an Intel processor. The first reason is that you have several levels of cache before the data even gets to the linux image. These layers (controller, z/VM, mdcache) will cut the time to retrieve the data, and caching it in linux as well really won't make anything any faster. The second reason is that larger images cause more paging on z/VM's part. The Intel memory suggestion is based on real memory, always available. Since this is virtual memory, your "in-memory cache" may actually be on disk anyway, causing a paging read when it is needed. You can actually slow down the response in an image by making its memory larger, because it will spend much more time paging. Finding the right size is a little more tricky. You want to start at some reasonable value (1G, 2G...), and then cut the image's memory over time until you notice it just begin to swap. Some people will allow it to do this minimal swapping, while others will increase the image size enough so that it doesn't swap at all. It just depends on which expert you choose to listen to. At this point, you'll have an image that supports the application running, but doesn't have enough memory to devote a large amount to caching its disk, which is exactly where you want to be. You may actually be surprised at how little memory this really is. -- Robert P. Nix Mayo Foundation.~. RO-OE-5-55 200 First Street SW/V\ 507-284-0844 Rochester, MN 55905 /( )\ -^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 2/4/10 7:00 AM, "Carson, Brad" wrote: > We have a project that is being setup to run on RHEL under z/VM. The sizing > parameters for the guests (Oracle DB, and WebLogic) are being sent to us based > on intel platform sizing. How do some of you handle the sizing conversion > from Intel to IFL's? Are there some rules of thumb, we should know about? > > Our environment: > > IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. > > > Thanks for any insight. > > /Brad (please ignore the company inserted HIPAA disclaimer) > > - This e-mail and any attachments may > contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If > you are not the intended recipient, any use or disclosure of this information > is STRICTLY PROHIBITED; you are requested to delete this e-mail and any > attachments, notify the sender immediately, and notify the LabCorp Privacy > Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. > > -- > For LINUX-390 subscribe / signoff / archive access instructions, > send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
First rule of thumb, don't guess. Since Oracle SGA can be anywhere from a few hundred megabytes to 10's of gigabytes, you need to size the server larger than the SGA. If you undersize the server, you WILL have undesirable performance. On intel they size based on low cost of storage and I/O avoidance so generally are too large. The SGA is usually oversized as well - BUT the oracle DBA will need to be involved as that is the person that can reduce the SGA size. There are ways to measure existing storage requirements, guessing without measurements is playing roulette (1 out of 38 you guess right?) Carson, Brad wrote: We have a project that is being setup to run on RHEL under z/VM. The sizing parameters for the guests (Oracle DB, and WebLogic) are being sent to us based on intel platform sizing. How do some of you handle the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? Our environment: IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. Thanks for any insight. /Brad (please ignore the company inserted HIPAA disclaimer) - This e-mail and any attachments may contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If you are not the intended recipient, any use or disclosure of this information is STRICTLY PROHIBITED; you are requested to delete this e-mail and any attachments, notify the sender immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
Good advice. Note, from experience, there is no way to compare apples to apples on zVM/Linux specs and Intel. We constantly get hammered by this when Managers cannot make valid cost evals when comparing the platforms. -Original Message- From: Linux on 390 Port [mailto:linux-...@vm.marist.edu] On Behalf Of Agblad Tore Sent: Thursday, February 04, 2010 8:09 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: A Question on Sizing z/VM Linux Guests just enough so it doesn't swap, then add 300K perhaps and two small (max 100k) virtual swapdisks and a larger 'real' swapdisk. In your case I would start with 2GB and see how it works. Cordialement / Vriendelijke Groeten / Best Regards / Med Vänliga Hälsningar Tore Agblad Volvo Information Technology Infrastructure Mainframe Design & Development SE-405 08, Gothenburg Sweden E-mail: tore.agb...@volvo.com http://www.volvo.com/volvoit/global/en-gb/ From: Linux on 390 Port [linux-...@vm.marist.edu] On Behalf Of Carson, Brad [cars...@labcorp.com] Sent: Thursday, February 04, 2010 14:00 To: LINUX-390@VM.MARIST.EDU Subject: A Question on Sizing z/VM Linux Guests We have a project that is being setup to run on RHEL under z/VM. The sizing parameters for the guests (Oracle DB, and WebLogic) are being sent to us based on intel platform sizing. How do some of you handle the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? Our environment: IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. Thanks for any insight. /Brad (please ignore the company inserted HIPAA disclaimer) - This e-mail and any attachments may contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If you are not the intended recipient, any use or disclosure of this information is STRICTLY PROHIBITED; you are requested to delete this e-mail and any attachments, notify the sender immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 - Please see the following link for the BlueCross BlueShield of Tennessee E-mail disclaimer: http://www.bcbst.com/email_disclaimer.shtm -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: A Question on Sizing z/VM Linux Guests
just enough so it doesn't swap, then add 300K perhaps and two small (max 100k) virtual swapdisks and a larger 'real' swapdisk. In your case I would start with 2GB and see how it works. Cordialement / Vriendelijke Groeten / Best Regards / Med Vänliga Hälsningar Tore Agblad Volvo Information Technology Infrastructure Mainframe Design & Development SE-405 08, Gothenburg Sweden E-mail: tore.agb...@volvo.com http://www.volvo.com/volvoit/global/en-gb/ From: Linux on 390 Port [linux-...@vm.marist.edu] On Behalf Of Carson, Brad [cars...@labcorp.com] Sent: Thursday, February 04, 2010 14:00 To: LINUX-390@VM.MARIST.EDU Subject: A Question on Sizing z/VM Linux Guests We have a project that is being setup to run on RHEL under z/VM. The sizing parameters for the guests (Oracle DB, and WebLogic) are being sent to us based on intel platform sizing. How do some of you handle the sizing conversion from Intel to IFL's? Are there some rules of thumb, we should know about? Our environment: IBM z10-BC (QA and Dev) and z10-EC (Prod) running z/VM 5.4 and RHEL 5.3. Thanks for any insight. /Brad (please ignore the company inserted HIPAA disclaimer) - This e-mail and any attachments may contain CONFIDENTIAL information, including PROTECTED HEALTH INFORMATION. If you are not the intended recipient, any use or disclosure of this information is STRICTLY PROHIBITED; you are requested to delete this e-mail and any attachments, notify the sender immediately, and notify the LabCorp Privacy Officer at privacyoffi...@labcorp.com or call (877) 23-HIPAA. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390