Problems on parallelizing...

Hi experts,
The question I'll mention afore hasn't been answered by our oracle DBA consultant... :(
So hope you can help me. May be I am not able to frame my question. So, Let me ask in this way.
1.- We launch a query against our database with empty tables,
2.- We fill the dables with data (about 30Gb),
3.- We launch the same query again.
Oracle takes a lots of CPU and when we get to the OEM we realise the query execution don't parallelize.
The query work properly but the times are indeed quite bad; in fact it seems Oracle get the SAME execution plan it used on the first time
and is unable to update it in a new way to optimize the new parallel query execution.
Neither us nor our DBA has been able to determinate yet what's happening, and why our system is parallelizing only in some cases and NOT always as we desire.
Technical specs:
We are running 11g over SuSE Enterprise Linux on a DELL PE 2900 III (May 2008) with 2 x Quad Core Intel Xeon X5450, 8 GB and 4 x 300GB SAS 15000rpm.
Hope you can help!
Noël

There are lots of variables here, even if it is a single table query.
For example :
1. DBMS_STATS has a parameter 'no_invalidate' which defaults to 'AUTO_INVALIDATE' which means that Oracle will not always invalidate existing execution plans for tables already present in the shared pool but will "slowly" invalidate them.
2. If the table contains Skewed data then the default method_opt=>'FOR ALL COLUMNS SIZE AUTO' , although expected to gather the right histograms doesn't always do so. IF you have high data skew, you might need to provide the specific columns.
3. DBMS_STATS uses a default estimate_percent=>'AUTO_SAMPLE_SIZE' which, may be an inadequate sample size, particularly if there is skew .
4. Data can be outlier values (eg you have 1 row that has a date '01-JAN-1900' and all other rows with dates '01-MAR-2009' to '23-MAR-2009', when Oracle does an cardinality estimate for range, it has to assume that some rows are between '01-JAN-1900' and '01-MAR-2009' !! -- estimate how many rows it thinks will NOT be in the MARCH-2009 !).
5. We are assuming that you have the right indexes which the Optimizer, also, on the basis of available statistics, decides are usable. If the statistics (for whatever sample_size was used) does not favour the index OR the index really isn't the right one, Oracle will do a FullTableScan.
6. Oh, and GATHER_TABLE_STATS defaults to 'AUTO' for the CASCADE parameter, meaning that Oracle might not gather index statistics, unless you explicitly specify CASCADE=>TRUE.
AND : Do you have the paramaters for Parallelism properly setup -- PARALLEL_MAX_SERVERS and other parameters and PGA_AGGREGATE_TARGET
Edited by: Hemant K Chitale on Mar 24, 2009 4:18 PM
Added note on parameters for Parallelism.

Similar Messages

SQL Developer video problems on Parallels (Win XP)

I am running ORACLE 11G, APEX and SQL Developer on my MacBook pro. Initially I installed 1.5.4 SQL Developer and I stared having problems when editing SQL scripts.
Edit window, blacks out the lines which are not being viewed. Very annoying when working on complex scripts.
Although I've reported this problem to Parallels support, given that SQL Developer is the only application causing the problem is making me think it is more likely an issue with SQL Developer.
Since then I went back to 1.5.1 version and having same problem.
Any body else having this problem?
I might try 1.5.4 with patch #2 . . .
Details related to SQL Developer installed:
Oracle SQL Developer 1.5.1
Version 1.5.1
Build MAIN-5440
Copyright © 2005,2008 Oracle. All Rights Reserved.
IDE Version: 11.1.1.0.22.49.35
Product ID: oracle.sqldeveloper
Product Version: 11.1.1.54.40
Version
Component     Version
=========     =======
CVS Version     Internal to Oracle SQL Developer (client-only)
Java(TM) Platform     1.5.0_06
Oracle IDE     1.5.1.54.40
Versioning Support     1.5.1.54.40
Edited by: APEX minded on 27/05/2009 17:12

I will post this info to help people who may have similar problem.
The work around I found is to run Parallels in full-screen mode.
Parallels support haven't provided a fix for this yet.

Wee Problem with Parallels Desktop and the Westmere Mac Pros

Just got my new Westmere 8-Core Mac Pro last Friday, and spent the last four days in agony.
I used Migration Assistant to transfer over my accounts from my old Mac Pro, but it crashed hard -- so hard that I had to reformat the hard drive and do a clean install of the system. So I decided to do a clean install of all of my programs as well, and the Westmere crashed on the Parallels Desktop installation. And, again, it crashed so hard that I had to reformat the hard drive and do a clean install of everything all over again.
Today, I read a review of the Westmere Mac Pros in Macworld (http://www.macworld.com/article/153704/2010/08/quad8_core_macpro_2010.html?lsrc=top2), and learned:
We ran into a problem running Parallels on the Westmere-based Mac Pros, so we could not complete the Speedmark 6 suite of tests. When booted into 64-bit mode (as the new Mac Pros do by default) the systems would crash with Parallels installed. Booting into 32-bit mode would allow the systems to start up, but we still couldn’t use Parallels. We eventually tried a complete wipe and reinstall from the Mac Pro’s systems disc and downloaded the very latest version of Parallels to eliminate any other potential problems, and had the same issue. Apple and Parallels are aware of the problem and assure me that Parallels is working diligently on a fix. The application worked just fine on the new quad-core Mac Pro with the Nehalem processor.
Hoped I saved somebody some time.

Hi akapod,
Looks like they may have the fix out.
Mac Pro (Mid 2010): Kernel panic when installing or restarting Parallels Desktop 5 for Mac
Please let me know if it's available and fixes it.
Thanks,
ivan

Problems with parallel compression

I'm having some problems utilizing parallel compression on some drums. Here's the setup:
-9 tracks of audio (mixed, EQ'ed, etc. ) outputting to a Bus, which serves as the master drum fader
-master drum fader has very light EQ and compression, and a limiter
-audio tracks also have a (post-fader) send going to a separate bus for parallel compression
-parallel comp bus is heavily compressed, slightly EQ'ed
-plugins used are all Logic's: limiter, compressor, channel EQ
The problem is that as I increase the parallel fader, the bass drum loses it's punch. It almost sounds like some kind of phase cancellation. The snare seems to lose a little presence as well. The cymbals, on the other hand, sound great. So if I leave the parallel fader very low, I'm not getting the compression I want out of the cymbals. Any ideas?
I'm still pretty new to mixing, and definitely to parallel compression, so it seems likely a problem with my plugin settings. I got some advice from a friend of mine, and tweaked a little from there, but here are the settings of the compressor on the parallel bus:
Threshold: -17 dB
Ratio: 10:1
Attack: 10 ms
Release: 11 ms
Knee: 0.0
Peak Mode
Thanks for your help.

I guess I'm just a little stuck on the usual parameters (ratio, threshold, attack/release, etc.), though I realize the Warmer doesn't operate in quite the same way.
Right. It is a bit confusing and there are only some controls which have usual plug-in equivalents. Look at VM as being a limiter, which in turn has an infinite ratio. The compression element of VM (as I understand it) affects the amount of compression underneath the ceiling. So the knee control affects that compression depth, rather than the compression curve, related to threshold, as seen in conventional compressors, like a sub threshold adjustment.
the knee sets the amount of compression applied to signals leading up to the ceiling, right?
Exactly. The ceiling, you would think, is the limiter ceiling for output a bit like Logic's adaptive limiter output. It is really a way of moving the output level to something other than 0 dbFS, and works with the saturation settings on the VM's back panel, and the way it handles the input side of VM. So I usually leave the ceiling at zero (as it's not being used for mastering) and change the drive, knee and saturation to taste. So the metering behavior you see, to me makes sense. Compare with pre and post, and adjust the drive, and it is more visually in step. I think it's more like a threshold than anything else. The knee is also affecting VM's make up gain, which you can balance with the drive knob to get the right amount of coloration.
Drive is, as you say, just an input level.
I always have FAT on. When used on drums, I change the rear saturation to less high and bass, more mid. I also change the freq. settings to be more active in the low mids, but this changes for each kit. I usually have brick wall and Auto (for release, not auto gain) on, with release set at X1.
The Mix control is mysterious to me. You would think that it works so that you can change the mix between an unprocessed input to the VM output and therefore use it on a drum submix without a separate bus for parallel compression, but it sounds different to me (and much better) when I use it on a bus set to 100%, so that's what I do, with some EQ before VM.

Has anyone had a problem running parallels 6.0 on a macbookpro?

I have a mid 2009 mac book pro. I have been using Parallels 6.0 for the last 6-9 months. Suddenly, in the last week, it doesnt start. I get the following error message
Make sure that the prl_disp_service process is active and it is not blocked by Firewall. For details, please refer to http://kb.parallels.com/8089. If the problem persists, contact the Parallels support team for assistance.tion via the link,
When I follow the instructions provided, I cannot run via terminal mode.

mike137137 wrote:
Thanks, BobTheFisherman.
Is there some way that I could have known it was there? Or that I could have found it? Searching on "parallels" found this thread, but not that one.
You can click on your name and see a list of your posts. Then click on the post of interest.
When you made that post you were in the Windows Compatility forum, that is why it is there.

Internet problems with Parallels.

Windows 7 shows that I have an active Internet connection, but I get the error "Your computer appears to be correctly configured, but the device or resource is not responding." My network settings for Parallels are "shared" and I've installed Parallels Tools.
I'm running the trial version of Parallels to see if it's worth buying. The internet worked just fine the first couple of days, but suddenly it won't load any webpages or even show my shared folders. I uninstalled and reinstalled Parallels but that didn't fix it either. I have no problems with my connection on the Mac side.
Any thoughts?

On Mac OS X navigate to Applications>Utilities and open terminal, then type line for line in the following.
$ su -
Password:
# echo > /Library/Parallels/.parallels_dhcp_leases
# killall -HUP pvsnatd
let us know if it worked

Problem in Parallel Dynamic Block

Hi All
I have a scenario in GP,
An initiator will make request and he/she want to get approval from multiple approvers. So the request want to go multiple users at a time.
Here I tried this with one Sequential block and a Parallel dynamic block.
In Sequential block I created inputform callable object using webdynpro DC and in Parallel dynamic block also created approval form using webdynpro DC.
Here i am facing a problem, while reaching the form to approval users, if any of the approver open it executed and submitted all other approver actions are also terminating or the inbox content is disapearing.
I want to exit from the Parallel dynamic block only after finishing all of the approvers submition.
Is any setting or modifications can be done to achieve this?
If anybody knows about this problem please reply me.
regards
Sajith

Hi Sajith
I think you are attaching multiple users to the same action (having approval form Callable Object). This will send the action to multiple approvals, but the process will be terminated on approval from any one of the multiple approvers. Due to this the request goes off from other approver's inbox.
I think for your scenario you can have a sequential block in which you pass the approver action to next approver only when the previous approver has approved it. You can do this by keeping a decision dialog process control, which gives you two actions approve and reject.
This will send the initiators request to multiple approvers and for the last approver you can terminate the process.
Parallel Dynamic Block is used for a scenario where you have to take a vote or survey, in this case you pass the survey document to multiple users (and multiple users are decided at runtime).
Please go thru this link to understand the Parallel Dynamic Block properly.
http://help.sap.com/saphelp_nw2004s/helpdata/en/43/7e5b306b3b2679e10000000a1553f7/frameset.htm
Regards,
Anish

Problem in Parallel Processing

Hi All,
i had performed all the steps involved in parallel processing like creating Multiline Container Elements
and creating Task.
when i run my workflow all the workitems are created in workflow logs in completed state
but in sbwp i can't find my workitems in the Inbox.
pls let me know why this is happening??
Thanks
Ravi Aswani

Hi Ravi,
As per as per your inputs, when you check workflow log all the workitems are in complete status.
1. Have you created all the tasks as background task?
if it is a background task, then you wont get any mails.
2. If you have created an foreground task and still if it is not sending any mails then check in the workflow log, if the flow is going to any activity step which is defined to send workitem.
3. Since you have said that the status is completed, i dont think there is any agent problem right now.
Please revert back if need more clarifications and with more inputs so that we can help you better.
Regards,
Gautham Paspala

Problem about parallel audit

Hello expert,
        I am using BPM on Netweaver CE7.1 EHP1. I meet a problems:
       There are 2 tasks: task1 and task2, and they are parallel task(The two tasks exist between parallel split and parallel join). If I run the task1 and return a parameter false, this means I needn't run the task2. But the task2 still exist in the owner of task2's UWL.I hope when someone complete task1 and return a parameter false, the task2 should remove from UWL.

Thank you for your replay .
   I explain my problem once more.
   Only the two tasks which are should launch at the same time all approve ,the process can complete. If task1 or task2 can not approve, the Process should relaunch. So I put Task1 and Task2 in the structure of parallel.My requirement is If Task1 return unapproved, the process should relaunch and the task2 should cancel in the UWL automatically.But problem is the Task2 did not be canceled.

Syncrhronization problems in Parallel Mode

I've developed a program im TestStand using the Parallel Process Model. I'm controlling 4 equipments with GPIB and a DaQ Card. When I run my first Socket I don't have problems and the sequence works perfect, but when I run more than one Socket the sequence fails and give me the error:"Attempt to unlock a lock that is not locked by this thread". Everytime I make some write or read operation to my equipment with GPIB I lock this step between lock and unlock. Can someboday help me? Maybe I'm not using properly the lock functions. Where can I find examples about this?

Without seeing your sequence it is difficult to say what exactly is going wrong. Basically, to use a TestStand lock, you create it with a name and lock it using the name before the critical section of your code. When you lock a lock, you specify a lifetime for which when it expires the lock will automatically be unlocked (I believe the default lifetime is until the sequence finishes executing). In addition you can perform an EarlyUnlock operation to release the lock before the lock lifetime expires.
Another way of using locks is to use the synchronization tab of the step properties dialog on any step. You give it a name of a lock and it will automatically create and lock the lock before calling the step's code module and then automatically unlock it after the step compl
etes.
-Doug

Problems with Parallels

The upgrade is working ok... for the first day, Parallels was fine (ok, 1 crash). Now, I can't open it!
What do I do?
Lindsay

Templeton Peck wrote:
Since Parallels is not an Apple product, you should post your question on Parallels own discussion boards:
http://forums.parallels.com
Definitely check out Parallels forums. But instead of posting right away, read the available information there. In the forums they have some FAQs and some personal experiences posted that will likely get you on the right track towards fixing the problem. At least that was my experience.

Few "minor" problems with Parallels, please help...

After successfully uninstalling Bootcamp and Windows XP, I bought Parallels for Mac and did okay on a clean install. The only problem is, everytime I plug in "ANY" USB device, I get a message (something to the effect) "Device being used by another process". When I attempt to "eject" that device, I still get that same message. Is there ANYTHING I can try??? I was hoping to either connect a flash drive with music and documents, or my 100 gig external USB hard drive, but Windows XP via Parallels does not recognize any external devices. By the way, I even went as far as installing the Mac drivers that Bootcamp requires. That did help with a few other "devices" (such as PCI devices, etc), but I cannot get Windows to recognize any other USB devices.
Your help, opinions and advice would be GREATLY appreciated.
Thanks
Lunceford

Since Parallels is not an Apple product, you'll get a better response by posting your question on the Parallels discussion boards rather than an Apple forum that focuses on Mac/Windows compatibility:
http://forums.parallels.com

Direct X Problem after Parallels Installation!

This is extremely urgent and any help would be apppreciated!
I was running Bootcamp with Windows Vista smoothly until I decided to install Parallels 4.0 on my Mac 10.5.8. After installing Parallels, it loaded my Bootcamp installation which sort of freaked me out and so I closed Parallels without doing too much. When I booted Vista today, I tried running games that requried Direct X or enhanced graphics and they are not working! These games used to work fine before I had installed Parallels! I even tried fixing the Bootcamp installation by repairing the drivers through the installation cd that came with Leopard, but I still get the same problem!
An example: I try running Project64, and it gives me a Direct X error 8876086A, stating that plugins are missing.
Some games crash back to desktop without any error.
*Please help!*

Hi and welcome to Discussions,
I admit I never allowed Parallels to use my BootCamp Windows, but this sounds to me like Parallels has somehow altered the state/version of DirectX in your BootCamp Windows, since Parallels only has support for DX 8.1 .
Have you run a 'DXDiag' to check the DX system on your Vista ?
Regards
Stefan

Problem with Parallels

So as part of my grad program, I have to play SIMCITY classic, which is provided for free online (I know...I think it's pretty ridiculous too!). I only own a macbook, so I recently purchased parallels and a copy of windows xp pro. SIMCITY online is only able to be played on a PC, so I thought this would fix the problem. However, now when I try to play it, it still doesn't work. It gets me past the screen that would normally tell me it's not compatible with Mac's, but then it just sits there with a blank screen. Any ideas of how I can make this work?

Oh, I'm sorry. I wasn't very clear in my original post. I'm not playing the game from a disk. It's played online. If you search for Simcity Classic Online, you can find the site. I'm pretty sure I can't post URL's here. So basically, its played right in firefox, or IE, or whatever browser I'm using. It just seems like there is something about the Mac that keeps it from working.

A possible problem with parallel operation?

Hello everyone,
         I am working on a labview program that uses parallel execution, and I am running into a possible problem. Please bear with me, as sometimes it is difficult to explain things happening in Labview in text alone. This application reads data from a server that we have in one of our labs via an ActiveX component. This server is known as a TIU server. The server contains various data including part numbers, serial numbers etc. It sends this data out in three different signals. A sequence count, an ASCII high, and an ASCII low. The sequence counter sends data in the following manner. When the sequence count is 0 and both ASCII high and low are NULL, then a new part number sequence is about to begin. When the count sequence changes to "1", then the next two string charaters listed under ASCII high and ASCII low represent a number that is a "Function ID". There are several of these, and each function ID contains different information either a part number, a serial number, etc. Once this is read, and it is determined that you are looking at the correct Function ID, the next Count shows how many charaters are to follow, and then you read each set of two ASCII characters and concatenate them together until you build the entire ASCII sequence as one variable and the sequence is complete.
       There are several equipment cards attached to this TIU server, and each card is free wheeling it's data out. As soon as power is applied, it will continue to send out all of its function ID's sequentially (Function ID 0-9) in order, in a circular fashion at a 1 Hz. rate. A really strange system to be sure, but it is a limitation that I have to deal with. So now that I have laid a bit of the background on this, here is the problem. Occasionally, (maybe one out of 20 trys), the data seems to get slightly corrupted. For example, in the enclosed code, I ran it 40 times, reading the same part number each time. In theory, it should always be the same result. But it is not. I am wondering if there is an issue with running the "READ TIU SERVER.vi" in PreAllocated clone reentrant execution mode? Perhaps on occasion, one of the instances does not fully close, before another instance begins, and a few bits get flipped around? I suppose it is possible that occasionally the TIU server is actually putting out erroneous data (and I am actually reading everything perfectly), but it is a fairly mature system, and I have not gone down that path yet. I have included the .vi that I am using to read the server data "test.vi", and the "Read TIU Server.vi" that is set to Preallocated clone reentrant execution. In this example, the "test.vi" is only a single instance, but in the actual program, I am running 20-30 of these, as Preallocated clone reentrant execution as well. As there are 20-30 cards on each TIU server to get information from. I am hoping that this makes sense, and that I have laid it out in a logical fashion. Sort of hard to explain in text sometimes. Thanks for the help....
Attachments:
test.vi ‏123 KB
Read Tiu Server.vi ‏15 KB

Hi ,
      Thanks for taking a look. You caught me on the shift register, I had not gotten around to cleaning up a few things. However the "elapsed time" functions acutally are there on purpose. Unfortunately, it takes a VERY long time to get one of those part numbers, as only two ASCII characters are moved accross the bus at a 1 Hz. rate. I orginally had all wait statements, but I discovered that when you stop a while loop, it will continue until the wait timer has expired. With 20+ of these functions in parallel, it was taking upwards of 30 seconds for the program to finally end. From what I was reading on the forums of others that had that problem, a "elapsed time" function can be broken through, while a standard "wait" statement cannot. Now when I press the "STOP" button, the program ends almost immediately.
     This is not a standard piece of equipment that I am using, but I think the explanation is of how it works is fairly accurate. Would you have a guess as to why every so often the data gets corrupted? Other than the bus itself is actually putting wrong data out there? I suppose I should check that as well, but due to the slow rate of the bus, and the tools that we have, I will have to sit and watch it for a long time... Thanks...
Dan

Has any experienced problems with Parallels 9 after OSX maverick 10.9.5?

After install OSX Maverick 10.9.5, Parallels 9 will not load..... any comments?

Hey....
Save yourself a lot of time and trouble..... buy the upgrade to 10....($50). I did this and "problem solved"..... I tried the Parallel's support before and they could not fix the problem on loading the original software... I had to use my own experience to get the product loaded.... I hope this problem does not present itself during other major upgrades....
Lots of luck...
wa4gwb

Problems on parallelizing...

Similar Messages

Maybe you are looking for