Parallel Calculation @ Essbase 6.5

Regarding version 6.5, if one user submits one heavy request (that includes attribute members), does Essbase use all CPUs? With 6.1.4 that we have here now, I see that it does not! We had once a spreadsheet request that needed more that 2,5 hours! In other words, does the Parallel Calculation of Version 6.5, include user requests? When does Essbase do parallel calculation and when it does not?As you understand, this relates to the first question of mine. (see "Best Hardware Server for Essbase" below) Thanks in advance,Antonis

No, it will not. The parallel calculation engine only works for calc scripts (e.g. when you build up your database). It does not parallel process user requests (e.g. pivots, zooms, retreivals). However, each request will pin a processor. So, your quad processor server will be pinned if 4 users simultaneously request any data.

Similar Messages

Essbase 64 bit on Windows 2003 Server: recomm. Hardware for Calcparallel ?

Hi,
1 year ago, we migrated to system 9.3.1 for Hyperion Planning - as usual, this includes Workspace, Financial Reporting, Financial Analysis, SmartView, Shared Services...
everything is running well on acceptable speed and in general 2-3 times faster than our old environment- essbase 7, planning 4 on (HP DL580 G3 windows 2003 32 bit servers, with each 4 Xeon monocores @3 Ghz)
we decided 1 year ago for 3 windows servers, each with 2 Xeon quattro-cores (=8 cores @ 3.0 Ghz)
1st server Financial Reporting + Workspace (Windows Server 2003 SP2, HP Proliant BL460C G1 with 2x Quattrocore Xeon E5450 3.0 Ghz) 8 GB RAM
2nd Server Planning+Shared Services (Windows Server 2003 SP2, HP Proliant BL460C G1 with 2x Quattrocore Xeon E5450 3.0 Ghz) 4 GB RAM (8 GB would have been better)
3rd Server 64 Bit Windows 2003 with Essbase (Windows Server 64 bit SP2, HP Proliant DL380 G5 with 2x Quattrocore Xeon X5450 3.0 Ghz) 16 GB RAM
as our planning application is quiet big and we run a lot consolidation & datacopy at night, we depend on parallel calculation.
We mainly use command calcparallel 4 - but on 64 bit system, calcparallel 7 should be possible as well
We plan now to extend the system for other regions/2nd application
My Question: the new consultant confused me today, that essbase would regard 1 physicial quattrocore processor as 1 core for parallel calculation - that a machine with just 2 physical cores is not a really powerful machine... and that we better think about new hardware- that a quattrocore is just 1 core for essbase analytics - is this true ???
so did we actually choose a weak environment with only 2 (Quattro)-cores) for Essbase ?
Can anybody confirm if there is a difference for essbase between having e.g. 4 Dualcores and 2 Quattrocores ? (Besides the fact, that it will cost more, as we have to lease a bigger Server with more processor-sockets- and that it will cost a lot efforts to get the system changed)
P.S.: EAS console shows 8 cores for the Essbase Server in properties
Applicationlogfile shows: calculating in parallel with [4] threads (we use command calcparallel 4 in businessrule)
- our nightly calculations take between 60 and 120 minutes- if we avoid fragmentation (run dense restructure every 4 weeks) and keep the system clean and lean...

Hi John,
thanks for the helpful link - the long discussion reminded me again, that pure cores and cache is not all, what makes up a good essbase application -
I will test with calcparallel 6 (in comparison to 4) on our 64bit environment - just to see if I can achieve a 10, 20 or 30% improvement...
In addition, I must admit, that 50% of our 1-2 hour calculation is calculated in serial/ not parallel... so I expect an additional 5 to 10 min improvement, not more.
Thank you for now

Essbase SET CALCPARALLEL 3 function

Guru's
I am trying to understand how CALCPARALLEL should behave in essbase when we have one environment with 2 processors, i did run one simple CALC ALL commnad to check parallel calculation with SET CALCPARALLEL 3 setting, and i can see in logs it says calculation in parallel with 3 threads, how's that possible while we have only 2 processors on this new box ?
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1013167)
Received Command [Calculate] from user [admin] using [tst.csc]
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012675)
Commit Blocks Interval for the calculation is [3000]
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012717)
Remote bitmap cache is [Disabled]
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012684)
Multiple bitmap mode calculator cache memory usage has a limit of [50000] bitmaps.
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012669)
Calculating [ Accounts(All members) Year(All members) Scenario(All members) Market(All members) Product(All members)]
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012678)
Calculating in parallel with [3] threads
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012679)
Calculation task schedule [5,2,1]
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012680)
Parallelizing using [1] task dimensions.
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012672)
Calculator Information Message: Completion Notice After Processing [11] Blocks
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012672)
Calculator Information Message: Completion Notice After Processing [11] Blocks
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012600)
Member [Profit_%] attempts to divide by Missing, Invalid, or Near Zero value (Message will not repeat)
[Thu Sep 13 18:16:46 2012]Local/Demo/Basic/admin/Info(1012579)
Total Calc Elapsed Time for [tst.csc] : [0.125] seconds

First of all (and I apologize if this is telling you things you already know) a simple count of processors does not necessarily tell you how many parallel threads your environment actually supports. It depends on the number of cores and processor architecture (hyperthreading etc).
The other point is that Essbase will happily split a parallel calc into more threads than are actually available at the hardware level. I just ran a calculation on my laptop which reported 8 parallel threads, but the hardware only supports two.

RAM based Essbase

Recently we have upgraded the hardware server to 96GB RAM for our Essbase database. Currently we have deployed some physical memory as the disk cache for the database. As our database size is just around 40GB only, can we have a way to put the whole database into physical memory for calculation, and then write back to disk?
We would like to know if someone has tried this approach to speed up the Essbase business rules parallel calculation, data load and data rollup, financial reporting, etc.
Thanks!

Some software vendor has provided RAMDISK software to enable to keep the whole database into physical memory. Under Windows 2003 Server Enterprise Edition, the RAMDISK can be up to 512GB!
http://www.superspeed.com/servers/ramdisk.php

Tuning Essbase Calc Scripts

Hi,
I am looking for some guidance on tuning a calculation script.The business rule perform forecasting for two years based on the historical data (upto 5 years). The rule performs well when it is executed against a single department, single activity and single project.However when I try to run the admin rule which performs the same function but for all departments, all classes and all projects. it never completes.
Your feedback is really appreciated.
Sincerely,
JJ
Note in this example:
Activity and Project are dimension.
Total# of Departments = 7000 members
Activity = 100 members
Project = 100 stored members
Accounts = 350 stored members

Hello JJ,
for tuning an essbase database a consultant may use a day or more. Not something to be handled properly in a forum.
What we can do here is to give you some guidelines.
Calculate only what is necessary. Data which have been calculated and are not changed, do not need to be recalculated. Data which will be modified in a next step, should not be aggregated and then be aggregated again.
When you calculate, much is related to to the block size and how fast Essbase can get the blocks into memory and back to disk. So optimize the blocks (dynamic calc on the parent members, label only on where no data storage should take place).
Then optimize the calculation. Work with parallel calculation. And test, test, and test.
Regards,
Philip Hulsebosch

AddIns("Hyperion Essbase OLAP Server DLL (non-Unicode)").Installed = False

I am using Hyperion Essbase 11.1.1.3 Version. Previously I was using Hyperion Essbase 9.3.1. I have used "AddIns("Hyperion Essbase OLAP Server DLL (non-Unicode)").Installed = False" for Essbase 9.3.1. Can anyone please let me know the similar VB code for Essbase 11.1.1.3?

For example, We have 3 Planning applications and when we run business rules on these 3 applications, Essbase divide his tasks using always 13% of CPU (1/8)...^^^Essbase is single threaded. Unless you use parallel calculation in the BRs (not that I particularly recommend that within the context of a Planning app) that's Working As Designed.
If you're curious about parallel calculation, the DBAG has plenty on it and the Tech Ref has all of the supporting commands and settings fully documented.
When parallel calculations work, they work really well. Sometimes there is zero (even negative) performance improvements.
Regards,
Cameron Lackpour

Forced serial calculation

I am using parallel calculation where possible. I havesome scripts that Essbase is forcing serial calculation on, and Ican't figure out why. I checked the documentation for thecriteria that will force serial calculation, and none of thecriteria apply. The message given is 'Formula on member[456-01-010-81-2001] forces calculation to execute in serial mode'. There is no formula on that member. It is not dependenton any othermember (nor do any of the other members have dependencies). There are no dynamic members involved in the calculation. I have identical calculation scipts (just with differentaccounts calculating) that calculate inparallel with no problem. Any help would begreatly appreciated. FIX(&CurrentYear,"Final","Actual",&CurrentMonth,"PostAlloc", @LEVMBRS(Customer,0),@REMOVE(@RELATIVE("AF_Less_SoldBus", 0),@LIST(@UDA(Product,"Inactive_Product"),@RELATIVE("Total_C&W",0)))) "456-01-010-81-2001"="456-01-010-81-2001"->"GLAlloc"->"SKU-Alloc"->"All Customers"* (Syst->"DP - Fixed Exp Cases"/syst->"DP -Fixed ExpCases"->"AF_Less_Sold_and_C&W"->"CustomerTotal"); ENDFIX

John mentions that you can set it in the cfg file and shows you the setting specs, but if you just want to do it for a single application and have it in the default calc, you also need to know how to set the default calc. I find the easiest way is to crreate a regular calc script with what I want the default to be. then validate it. Out of the box, the default calculation is nothing more than a Calc All; statement. It can be anything you want. Once you have created the calc script. The following link tells you different methods for setting it
http://docs.oracle.com/cd/E17236_01/epm.1112/esb_dbag/dcaintro.html#dcaintro507
and If you want to do it from EAS look at http://docs.oracle.com/cd/E17236_01/epm.1112/eas_help/defcalc.html
It is important to know that if you change the calculation script you used to create the default calculation, it does not change the default calc itself, you would have to reset it.

Enabling Parallel Calc (CALCTASKDIMS)

Hi,I'm looking for information related to the CALCTASKDIMS function. The issue is to identify additional tasks for parallel calculation. As I rode it in DBA guide, Essbase uses the last spare dimension in an outline to identify tasks that can be performed concurrently. But how to know if I need to specify calctaskdms settings a second or a third sparse dimension???Thanks for your helpRegards,Sébastien ROUX

I saw this in DBAGThe 50% empty task ratio pointed out probably is a good goal of tuning with CALCTASKDIMS ? Use this configuration setting only if your outline generates many empty tasks, thus reducing opportunities for parallel calculation. See the Essbase Database Administrator's Guide for more information about what kind of outlines or calculation scripts generate many empty tasks. ? Essbase writes a message to the application log specifying the number of tasksthat can be executed concurrently at any one time (based on the data, not thevalue of CALCPARALLEL or SETCALCPARALLEL):Calculation task schedule [576,35,14,3,2,1]This example message indicates that 576 tasks can be executed concurrently.After these tasks complete, 35 more can be performed concurrently, and so on.The benefits of parallel calculation is greatest in the first few steps, and thenthe benefits taper off because there are fewer and fewer tasks being performedconcurrently.? Essbase writes a message to the application log indicating how many of thetasks are empty (contain no calculations):[Tue Nov 27 12:30:44 2001]Local/CCDemo/Finance/essexer/Info(1012681) Empty tasks [91,1,0,0]In the example log message above, Essbase indicates that 91 of the tasks atlevel zero were empty.If the ratio of empty tasks to the tasks specified in the task schedule is greaterthan 50, then parallelism may not be giving you improved performance,because of the high sparsity in the data model.Steven [email protected]

Essbase Performance Issues on Sun Solaris 10

We have a new Hyperion Environment 11.1.1.3 with Essbase sitting on a Solaris box. We are running a calculation script under the "FINSTMT" database that is called CALCALL. This is the default calculation for a database in Essbase (it runs a command called CALC ALL). We are running this same calc against the same database outline and data set across the environments to benchmark performance.
The script in the new environment should run faster, but it runs slower. The server is basically sleeping and we were curious if anyone can recommend configurations within the app or for the OS? Things like semiphors, shared memory, etc... Also if anyone has suggests or ideals to tweak Essbase performance on a Solaris 10 machine and/or UNIX. What should I do to the Essbase.config file?
Mike

I can't help you solaris tuning, but some things to look at.
1. Is the Essbase.cfg file the same on both servers? You might have parallel calculation turned on in one and not the other. Caches could also be set differently
2. Are the database caches set the same? This could impact performance as well
3. Are you doing an apples to apples comparison? Is one database loaded and recalculated many times while the other is not (or restructured or reloaded)

Performance issue - Loading and Calculating

Hi, I am having 5 GB of data. It is taking 1 hr to load and 30 min to calculate. I did the following things to improve the performance. 1) Sort the data and loading them in the order of largest sparse first, followed by smallest and dense 2) Enabled parallel load, gave 6 threads for prepare and 4 for writing. 3) Increased data file cache as 400MB and data cache as 50MB, then index cache as 100MB. 4) Calculation only for 4 dimensions, out of 9. In that 2 are dense and 2 are sparse. 5) Calculation with parallel calculation having 3 threads and CALCTASKDIMS as 2. But i am not getting any improvements. While doing the calculation i got following message in the logs. I feel that CALCTASKDIM is not working [Fri Jan 6 22:01:54 2006]Local/tcm2006/tcm2006/biraprd/Info(1012679) Calculation task schedule [2870,173,33,10,4,1] [Fri Jan 6 22:01:54 2006]Local/tcm2006/tcm2006/biraprd/Info(1012680) Parallelizing using [1] task dimensions. Usage of Calculator cache caused reduction in task dimensions [Fri Jan 6 22:33:54 2006]Local/tcm2006/tcm2006/biraprd/Info(1012681) Empty tasks [2434,115,24,10,2,0] Can any one help me what the above log message is telling and what are the other things to be done to imrpove the performance. Regards prsan 

its not the problem with ur calc task dim. Calculation task schedule [2870,173,33,10,4,1] indicates that ur parell calc can start with 2870calculations in parallel, after which 173 can be performed inparallel then 33 ,10,4 and 1. Empty tasks [2434,115,24,10,2,0] means these manytasks dont need any calculation- either because there is no data orthey are marked clean due to intelligent calc. the problem lies with your cal cache setting. try increaing thecal cache settings in ur cfg file and use calcache high setting inyour calc. hope this works

Dynamic Calc processor cannot lock more than [100] ESM blocks during the calculation, please increase CalcLockBlock setting and then retry(a small data cache setting could also cause this problem, please check the data cache size setting).

Hi,
Our Environment is Essbase 11.1.2.2 and working on Essbase EAS and Shared Services components.One of our user tried to run the Cal Script of one Application and faced this error.
Dynamic Calc processor cannot lock more than [100] ESM blocks during the calculation, please increase CalcLockBlock setting and then retry(a small data cache setting could also cause this problem, please check the data cache size setting).
I have done some Google and found that we need to add something in Essbase.cfg file like below.
1012704 Dynamic Calc processor cannot lock more than number ESM blocks during the calculation, please increase CalcLockBlock setting and then retry (a small data cache setting could also cause this problem, please check the data cache size setting).
Possible Problems
Analytic Services could not lock enough blocks to perform the calculation.
Possible Solutions
Increase the number of blocks that Analytic Services can allocate for a calculation:
Set the maximum number of blocks that Analytic Services can allocate to at least 500.
If you do not have an $ARBORPATH/bin/essbase.cfg file on the server computer, create one using a text editor.
In the essbase.cfg file on the server computer, set CALCLOCKBLOCKHIGH to 500.
Stop and restart Analytic Server.
Add the SET LOCKBLOCK HIGH command to the beginning of the calculation script.
Set the data cache large enough to hold all the blocks specified in the CALCLOCKBLOCKHIGH setting.
Determine the block size.
Set the data catche size.
Actually in our Server Config file(essbase.cfg) we dont have below data added.
CalcLockBlockHigh 2000
CalcLockBlockDefault 200
CalcLockBlocklow 50
So my doubt is if we edit the Essbase.cfg file and add the above settings and restart the services will it work? and if so why should we change the Server config file if the problem is with one application Cal Script. Please guide me how to proceed.
Regards,
Naveen

Your calculation needs to hold more blocks in memory than your current set up allows.
From the docs (quoting so I don't have to write it, not to be a smarta***:
CALCLOCKBLOCK specifies the number of blocks that can be fixed at each level of the SET LOCKBLOCK HIGH | DEFAULT | LOW calculation script command.
When a block is calculated, Essbase fixes (gets addressability to) the block along with the blocks containing its children. Essbase calculates the block and then releases it along with the blocks containing its children. By default, Essbase allows up to 100 blocks to be fixed concurrently when calculating a block. This is sufficient for most database calculations. However, you may want to set a number higher than 100 if you are consolidating very large numbers of children in a formula calculation. This ensures that Essbase can fix all the required blocks when calculating a data block and that performance will not be impaired.
Example
If the essbase.cfg file contains the following settings:
CALCLOCKBLOCKHIGH 500 CALCLOCKBLOCKDEFAULT 200 CALCLOCKBLOCKLOW 50
then you can use the following SET LOCKBLOCK setting commands in a calculation script:
SET LOCKBLOCK HIGH;
means that Essbase can fix up to 500 data blocks when calculating one block.
Support doc is saying to change your config file so those settings can be made available for any calc script to use.
On a side note, if this was working previously and now isn't then it is worth investigating if this is simply due to standard growth or a recent change that has made an unexpected significant impact.

Calculating in Serial

I have a Business Rule create block script that is executing as a serial calculation, even though I have parallel enabled in the config file, and other rollup calculations on the same database execute in Parallel. Below is the script causing the problem:
Parameters
Member Run_Version Of "Version";
Member Time_Period Of "Time"
EndParameters;
create blocks {@LEVMBRS("Ledger",0) AND @IDESCENDANTS("Ledger"),[Run_Version],@LEVMBRS("Function Level",0) AND @IDESCENDANTS("Function Level"),@LEVMBRS("Geo Market",0) AND @IDESCENDANTS("Geo Market"),[Time_Period],"No Product","No Subaccount",@LEVMBRS("Department",0) AND @IDESCENDANTS("GROSS EXPENSES")};
VAR _HAMPrfx1;
SET UPDATECALC ON;
FIX([Time_Period])
FIX((@LEVMBRS("Ledger", 0)) AND (@IDESCENDANTS("Ledger")))
FIX([Run_Version])
FIX((@LEVMBRS("Function Level", 0)) AND (@IDESCENDANTS("Function Level")))
FIX((@LEVMBRS("Geo Market", 0)) AND (@IDESCENDANTS("Geo Market")))
FIX("No Subaccount")
FIX((@LEVMBRS("Department", 0)) AND (@IDESCENDANTS("GROSS EXPENSES")))
"No Product" (
@CALCMODE(TOPDOWN);
_HAMPrfx1 = "No Product";
"No Product" = 0;
"No Product" = _HAMPrfx1;
ENDFIX
ENDFIX
ENDFIX
ENDFIX
ENDFIX
ENDFIX
ENDFIX
And here is the log file:
[Tue Mar 10 15:52:25 2009]Local/Allocate/AllocTS/CJN013/Info(1012571)
Presence of variables or formulas with @XREF function forces calculation to execute in serial mode
[Tue Mar 10 15:52:25 2009]Local/Allocate/AllocTS/CJN013/Info(1012668)
Calculating [ Product(No Product)] with fixed members [Ledger(GAAP_Adj, Cash, Non-Cash, Allocation); Version(Revision1); Function Level(UW1, UW3, UW6, UW7, LAE1, LAE6, LAE7, INV, RE, No Function); Geo Market(GA, IL, IN, OH, WI, IA, KS, MO, MN, ND, NE, ]
[Tue Mar 10 15:52:25 2009]Local/Allocate/AllocTS/CJN013/Info(1012677)
Calculating in serial
Which is also witness by watching the server during the calculation... I'd like to make this step as efficient as possible.
My outline does not have any @XREF functions in it, or any cross database functionality.
Thanks

Graphical business rules -- I have to say that I don't do much with those because of performance issues/I am stubborn and like to roll my own.
Sorry I can't be of more help other than recommending using the Enhanced Calc Script functionality. You know, as you optimize the code, you may find that performance is perfectly acceptable without parallel calculation.
BTW, I more likely than not would not use the approach you highlighted to create a block were I writing the code directly.
Some of the approaches I might try?
1) DATACOPY
2) CREATENONMISSINGBLK (be careful)
3) CREATEBLOCKONEQUATION (I've got the name wrong on that one but it's in the Tech Ref)
Of course they blow away your requirement for a graphical approach.
Perhaps someone more versed in that tool can chime in.
Sorry I can't help more with that part of it.
Regards,
Cameron Lackpour

Reg:using dynamic calc in calculation script

Can we use a dyanmic calc member of sparse dimension in a calculation script. In which situations we use the intelligent calculation and in which situations we will not use intelligent calculation.

Generally you turn off intelligent calc for any manual calculation scripts you want to run. Intelligent calculation means that "If you then load a subset of data, on subsequent calculations, Essbase calculates only the data blocks that have not been calculated and the calculated blocks that require recalculation because of new data".
Taken straight from: http://download.oracle.com/docs/cd/E12825_01/epm.111/esb_dbag/dcaoptic.htm

BSO Calculation

I have 12 dimensions in my cube. I know it is not recommended to have these many dimensions but I am helpless against requirements. Now I have member formulas to do currency conversion in the dimension called Currency Type. The default calc takes 4 hours and because of formulas I cannot utilize the parallel calculation.
So what I did is
CALC DIM ("Currency Type");
CALC ALL EXCEPT DIM("Currency Type");
Will the above calc script provide me the same results as
CALC All;
Thanks

From an aggregation standpoint, it will most likely give you the same results, but not necessarily. You do have to consider order of dimension aggregation possibly giving you different results.
In a CALC ALL, dimensions are calculated in the following order:
1) Dimension tagged as Account
2) Dimension tagged as Time
3) Remaining Dense dimensions in outline order
4) Remaining Sparse dimensions in outline order
In your case, you're going to get something slightly different that could result in different numbers (particularly if you've got two-pass members). Here's what your order would be:
1) "Currency Type" dimension
2) Dimension tagged as Account
3) Dimension tagged as Time
4) Remaining Dense dimensions in outline order
5) Remaining Sparse dimensions in outline order
I guess I'm confused on why you're doing this, though. If you're looking for efficiency in your calc, I'm going to assume your dense dimensions are dynamic at all upper levels (and on the formula members), you don't have formulas on sparse dimensions, and all you need to do after converting currencies is to aggregate your sparse dimensions. If all that's true, use the AGG command on your sparse dimensions instead of CALC DIM or CALC ALL EXCEPT DIM.
Hope that helps,
Edward, Oracle ACE Director - Hyperion
http://looksmarter.blogspot.com/

Upgrade to Essbase 6.5.x

We're currently running 6.1.2 and want to upgrade to 6.5.x. Is there a recommended version? Are there any significant issues to be aware of?

We are using version 6.5.4 and found some bugs, like:- Getting data from nowhere while using parallel calculation (like fixing in a currency and getting results in other currency)- Getting #mi in a report with attributes and two-pass calc members- Not getting any value using a -> operator and a dynamic calc memberI was told that all this bugs are addressed in version 6.5.5, but I have not been able to test it.Anyway, I suggest you to check the bugs in the 6.5.4 and 6.5.5 readme files, to decide which version is more convenient to you

Parallel Calculation @ Essbase 6.5

Similar Messages

Maybe you are looking for