First, Thank You Jeremy.....
Ok, first question. Are your service levels even partially important? If you go production and you have bad performance and can not figure out what is wrong, say for 2-3 days, does anybody care? Most production installations have service level requirements, if you do not, then no need to pay for some insurance to try and ensure you meet those, and stop reading..... Two MAJOR Requirements: 1) You have the information needed, 2) Turning off the monitor does not solve your performance problem. This is not being silly, there are many agents and monitors you can run that will consume more resource that you would ever want to utilize. I like the following quote about one of those agents: "a single java client app from one of our agents when idle made my guest use 10 times as much cycles, 50% more disk space, and twice the memory" The monitors provided by Velocity Software are designed to use the least amount of resource and provide all the data you could need. In the last couple of weeks, every single positive non-vendor comment on the LISTSERV's have said to use Velocity Software products, for a reason!!! I see performance problems in many areas. I've seen some really good systems people have their system die because something silly was overlooked. If you care about meeting service levels, then you must resolve problems quickly in the following areas: VM: This includes DASD, Paging, Spooling, DASD Cache, Processor, Channels (ESCON/FICON/Fibre), 2GB Storage, >2GB Storage, Expanded Storage SCSI, QDIO, VSWitch, Diagnose counts, and MDC (and maybe a few hundred more VM things). Linux: Process data, Swap, storage, processor, Disk, and more. Network: Traffic, errors, etc.... And after the normal system measurements are done, look at the applicaton, Tom's comment is absolutely correct: "But the multiuser applications (Websphere and Oracle) are where the resources will be used. And that needs monitoring." So there are two requirements, 1) MEASURE, AND 2) MEASURE AT A REASONABLE COST IN RESOURCE.... Oracle: Storage, CPU, lots lots more. Websphere - watch out for the Java App that measures this. Risky to run in a shared resource environment. BUT, watch this space for what ESALPS provides, this is the current project due out this quarter. *** ESALPS *** Provides REAL TIME and HISTORY for the above VM data, Linux data, Network data, and has significant data to help you with Oracle and Websphere with more to come - at both a financial and operational price that most installations seem to appreciate.... ESALPS from Velocity Software includes ESAMAP, ESAMON, ESATCP and ESAWEB for those of you that just have one or two products. Or, there is zMON for those that want to keep the "$" low. And you could check out zTUNE, our new service to ensure you provide the optimal performance service levels, given your requirements. zTUNE provides the old ESATUNE functions that have helped VM installations for the last 10 years or so, and ** NOW INCLUDES LINUX **, Network and much more. And is more flexible so that checks can be easily added for much more. "HTTP://VelocitySoftware.com/ESALPS.HTML"; "HTTP://VelocitySoftware.com/zqTUNE.HTML"; So, Too meet your service levels? or not.... That would be the question. /advertisement.... >Date: Tue, 1 Feb 2005 13:16:25 -0600 >From: Tom Duerbusch <[EMAIL PROTECTED]> > >Yep, under VM. > >And yep, I know about CPU inaccuracties...have/had the same problem >with VSE and CICS for decades. > >But if VM says a machine is using 50%.... >And in that machine, TOP says some process is the major user... >Then, you look at that process. > >Don't need fancy tools for that. > >But if the fancy tools then say, this process is using xx service, to >reduce the cost of this service, tune yy, then it's a good deal. > >I assume that there is some sort of Oracle performance monitor and >Websphere performance monitor to say what is actually going on. But >from the Linux perspective, just what would it tell me? > >Just weighing the decision....money for Oracle performance tools, vs >money for Linux performance tools.. If money was no object, I would get >the VM performance, Linux performance, TCP/IP performance, Oracle >performance, Websphere performance tools. > >But in real life, I will have to rank them. The bottom ones may get >cut out. > >I would rather wait to see where the performance problems hit with the >new workloads, but what makes it in the Capital budget, doesn't come >out of our Operational budget later. > >Tom Duerbusch >THD Consulting > >>>> [EMAIL PROTECTED] 02/01/05 1:03 PM >>> >You may want to search the archives for some of Barton Robinson's very >detailed messages regarding how top,vmstat,etc provide inaccurate >(sometimes VERY inaccurate) information when running under VM. We >have >been very happy with the ESAMON tools and what it tells us. It >measures >darn near everything you can measure, so when you need to look >someplace >you never looked before chances are you are measuring something.. Plus >it >normalizes the data to give you a more realistic view of the workload. > > >----------------------------------- >Jeremy Warren > >Tom Duerbusch <[EMAIL PROTECTED]> >Sent by: Linux on 390 Port <[email protected]> >02/01/2005 01:55 PM > >I know there are Linux performance measurement tools out there, some >free, some at a cost. > >But a more basic item. Just what needs to be meaured? > >We may be increasing our Linux workload greatly. We have been using >Linux (Suse) for about 3 years now. So far, I've been satisified with >what I get from VMSTAT and TOP. Of course, we haven't had many >problems and when we have had problems, they were not tramatic. > >Now, we may be moving to mission critical. > >The real work is going to be done in Oracle and some in Websphere. It >seems to me that once you set them up properly in z/Linux, the >monitoring and measurement must be done within those applications. > >So, what do I need to know in the Linux machine when 98+% of the work >within that machine is caused by a single "big" application (Oracle >or Websphere)? > >Tom Duerbusch >THD Consulting "If you can't measure it, I'm Just NOT interested!"(tm) /************************************************************/ Barton Robinson - CBW Internet: [EMAIL PROTECTED] Velocity Software, Inc Mailing Address: 196-D Castro Street P.O. Box 390640 Mountain View, CA 94041 Mountain View, CA 94039-0640 VM Performance Hotline: 650-964-8867 Fax: 650-964-9012 Web Page: WWW.VELOCITY-SOFTWARE.COM /************************************************************/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
