BIA question...

Hi,
I have activated the BIA index for some of our cubes for our testing purpose. Now if BIA is active for a cube the query first checkes if data for report is available in OLAP, then it check if BIA is active for this cube and at last it reads from Cube if both are not available.
How can I make sure that the query has used BIA index?
Is there is any table where it writes a staistics information?
Please let me know if anybody has any idea about this.
Thank you everybody in advance.
SC

Hi,
since all queries read the data from the BIA - if the BIA index is active for this InfoCube, there is no "real" need to check it.
But if you neverthelesse want to check it, you should go to the Query runtime statistic tables or the appropriate technical content.
The mentioned table RSDDSTATTREX holds the statistics for the BIA-index fill times, not the queries.
The table RSDDSTAT is no longer valid for the query/OLAP statistics. The new statistics information can be found in the tables/views:
- RSDDSTAT_OLAP for all Frontend and OLAP times, and
- RSDDSTAT_DM for all Datamanager times, i.e. the data read access times, either on the DB, the BIA, or a remote system if you use a virtual InfoProvider.
If you then execute a query, that reads data from the BIA server, go to SE16 and get the STEPUID of your query execution in table RSDDSTAT_OLAP. With this STEPUID you can select in table RSDDSTAT_DM the corresponding entries of the read access. Now in these rows you find in the column "aggregate" the entry <InfoCubename>$T if the data was read from the BIA server (techn. name of the BIA index). If the data was read from an aggregate you find the technical name of the aggregate, somthing like 1000xx there, or the name of the InfoCube.
Btw: make sure that you have turned on the statistics for the InfoProvider or the Query and have set the correct detail level(transaction RSDDSTAT) otherwise you will not see what you want.
You can also use transaction RSRT1 and the debug-option "show statistics". After you see the query result, press F3 and you see two Tabs with the OLAP/FE and the Datamanager statistics.
Of course you can do the whole analysis much easier using the technical content for the query runtime statistics, but this here can be used as a quick check.
Hope this helps,
Klaus

Similar Messages

Accessing BIA through Business Objects - read from Query or InfoCube

Hi All,
I've a BObj/BIA question, some similar threads have been posted....
With a Business Objects front-end, I understand that BIA can be used to access Indexed InfoCube data, and reduce the data manager portion of the overall runtime.
My question relates to the options of a)Business Objects reading data defined in Queries, or b)BO reading the data from the InfoCubes.
So my question is:
a) Is there any difference/benefit in using one option over the other? (queries Vs InfoCubes directly)
b) Will both methods access the required data through the BIA if it (the data) is available there and indexed?
Thanks for the help,
Mk.

Hello Michael,
both methods will use BIA. We recommend to use queries since they typically limit the selection criteria through variables and give you the possibility to use restricted and calculated key figures. With direct InfoCube access you run the risk that users try to read the complete InfoCube.
Regards,
Marc
SAP NetWeaver RIG

Question- Colgate Palmolive,, master data maintenance in BIA

Hi BW GURU, Like You said master data in BIA gets updated, when we run change run and it actually considers BIA index for a cube on BIA just like another Aggregate Right?
So Question,,, when we run change run it brings all changes into F table of an aggregate and then updates the records and then compress it (back to E table) and then after updating all aggregates it updates the master records. so when it updates the changes in BIA and how can we monitor those during change run?
Thanks.
Faisal

Hi Marc Thanks for Your reply.I understand now that Change run updates all master records in BIA as well upon execution of change run.
I have an other question.
Suppose we have 10 blades on BIA. We have a cube with 10 M records, when it creates BIA index for this cube it will store data like buckets on BIA disk ,,lets say it created 10 buckets, each had 1 M records. and then loads it to BIA main memory.
What if We had say 12 M records. Will it still create indices in buckets by keeping in mind that it has to distribute data evenly on every blade upon loading index for this cube into main memory?
Regards
Faisal

BIA Reorg Question

Hi,
This is for BIA at Rev.41 on HP Blades.
When I go to TREX Administration GUI on UNIX and then go to Reorg tab then I see that the Traffic light in Reorg column in line "Standard Deviation of all Indexes over IndexServers" is "Grey n.a.". I was expecting it to be green no or yes.
The deviation is over 40% now and still no red alert for Reorg which is not normal to me. I am comming from IBM Blades environment and it normally generates a red alert around 20% (do not remember exact %).
Is it normal behaviour?
The other lines are ok.
Is there any please we can define the threshold or it is pure technical. If it is then what is a standard threshold for HP.
Please let me know if there is any config setting is missing.
Regards.
Sume

Thank you MIke for the reply.
I do not think you have understood my question completely.
I am not worried too much about the algorithim about thresholds, even though I have asked about it.
My main question question is that I should not see "<Grey Icon> n.a." in the column Reorg in TREX Admin. GUI at UNIX.
This means that there will not be a Reorg red alert even if the deviation 100% and it is true because I have not seen a red alert even after deviation was 98%. The red alert for Reorg is generated based other lines. There are total four lines.
Please look at the GUI and you will understand it.
I am comming from IBM shop and I have not seen this on IBM Blades. It either shows Red or Green with Yes or No. It is never n.a. That was Rev 38.
This is Rev 41.
I thnk this is the issue unless SAP says that this is the way it is and changed in Revisions after 38. I do not think it has anything to Vendor.
Please Rectify.
Regards.
Sume

Error while deleting from BIA index

HI EXPERTS GETTING BELOW ERROR PLZ GIVE CLARITY ON THIS:
1.A communication error occured, with the TREX TcpIp
2.Error while deleting from BIA index; reconstruction required
THANKS &REGARDS,
rAMESH,

Dear Ramesh,
Please check the connection with BIA. Please ask the Basis team to check the BW to BIA connection if they maintain it.
YOu can go to rsddbiamon2 and check BIA connection availibility for more details.
If that is right please try redoing the step at which you faced error and it should help.
Please close the question if you have got the answer or solved it.
Regards,
Den

BIA Index Loading into Mmeory (for SAP) ....

Hello all,
I went to www.sdn.sap.com and chose Business Intelligence from the left menu.
In the BI Knowledge Center (Right side), I chose BI Accelerator, which is under Key Topics section.
Then I opened a document "SAP Business Intelligence Accelerator (PDF 154 KB)".
Go to the section "SAP BI Accelerator at Work", which is at page# 5. The third point says like this -
BI accelerator indexes are loaded into memory where the query is processed. In memory, joins and aggregations are done at run time. Loading of indexes into memory happens automatically at first query request, or it can be set for preloading whenever new data is loaded.
I had an understanding that it will be loaded into memory only when query was first executed. But this says it can be set for preloading as well wherever new data loaded.
My question is that where this setting can be done for preloading?
It says that preloading option is available whenever new data loaded so does that mean that it can only be set for rollup of BIA Index and for initial load, it will still load data into memory when the query is first executed.
I would appreciate if somebody has knowledge of this as it is new technology.
Others, who has a knowledge can also answer.
Your help will be greatly appreciated.
Thank you to everybody in advance.
Sume

Hi,
I found it.
It is an option on BIA Index Property button in the wizard. It appears only after the initial load is done. There is check box for to keep the BIA Index in main store. I thin it is only applicable to Roll-up.
Thank you.
Sume

BIA running under VMware

Please don't flame me as I am just trying to understand licensing, scalability, and cost of this solution. My understanding of the BIA appliance consist of Linux blades (16GB of memory), share SAN disk subsystem (GPFS configured), isolated switched infrastructure, and the application software (TREX). Speaking to the physical configuration, this could all be virtualized into a custom VMware configuration where you could maintain/guarantee a high level of performance for disk, network, and Linux guest/blade performance. Now for some of my questions.
My understanding of the license is per 16GB where 16GB is the max per blade. This license is for production usage and would be priced different in a dev/test environment. Is this correct? To move a dev blade to a production blade center would have a license cost, correct? Also, if this solution was virtualized, an additional license would be needed when adding new virtual Linux guests to the solution.
Based on licensing, would it be possible to virtualize the dev/test environment for no or little cost and concentrate the hardware solution to production usage?
What have people done when they max out their current blade center (10 or 14 depending on HP or IBM)? Just buy another appliance? It would see easier to support a configuration where you could add a VMware guest and just purchase the additional software license. A win for SAP and the customer! This would allow an easier scalability path when loads increase. I haven't heard/read about processor being a bottleneck so I have left this out but please let me know if this is an issue.
I know this is easier said then done but with virtual being supported in all other areas of the SAP landscape, it would make sense to allow it to creep into the BIA solution.
What do you guys think? Am I crazy for even thinking it possible?
Thanks,
Steve

Hello Steve,
first about licensing... There are no SAP license fees for the non-productive blades. If you move blades from dev/qa system to production, then you new license fees are due. Please talk to your SAP account executive in this case.
Next about virtualization... BI Accelerator is all about performance. BIA is highly optimized for parallel processing and taking advantage of Intel CPUs. VMware or other attempts to virtualize BIA will not get even close to the real deal. It's basically a waste of time to even try.
Nevertheless there is one case where it makes sense. For a training system you would not need good performance and you would be using the "virtual BIA" only with a single training user. We have been thinking about it but as you said, it's not as easy as it sounds.
Regards,
Marc
SAP NetWeaver RIG

How to check if BIA has effect on a BPS function

Dear Experts,
As I know BIA can also improve DB read time (data selection) when performing a BPS function. Currently we are using a BIA test installation to see if it will help us. We have some BPS function with long data selection times.
I know created BIA indexes on all involved real time cubes. By just executing the concerned BPS functions I can see no effect. I now want to check if the BPS functions do access the BIA at all. For queries this is quite easy to check with RSRT.
So my question is:
Is there any tool available to check wether BIA is used by a BPS function or not? Some kind of RSRT for BPS.
Thanks & Regards,
Ulrich Meier

Hi Jens,
Thanks. I tried this. Strange result is that I get no data in that trace when running a BPS function. I do get data when running that trace during execution of a query which runs on the same InfoCubes as the planning function.
In the meantime I found table RSDDSTATBIAUSE which gets updated each time a InfoProvider is accessed if it has a BIA Index. Here my InfoProviders get a +1 in field BIA_USED when accessing them with the BPS function.
Furthermore I came across OSS-Note 990000 and installed the latest version of report ZBPPOBPS70 (I already knew that from BW 3.5). Here I found that field 'BW:Used Aggregate' is not populated, although within the coding of ZBPPOBPS70 there is indication that it should either show the used aggregate or BIA Index. The data is read from table RSDDSTATDM field AGGREGATE. There I found entries in AGGREGATE for 'normal' query calls but none for BPS calls.
With alll that in mind I'm more and more confused...
Regards,
Ulrich

How to stop a BIA query

Hi,
Sometimes my users run a BIA query by mistake or a hug BIA query that running too long, they want Basis to stop/kill their query sessions. They start the query from the Portal ->BI ->BIA. Given that they provide me user-id, I can go to the BI and kill their RFC sesssions. That would kill their RFC sesssions in the BI.
My question, how do I kill their sessions from the BIA so we can avoid memory over loaded? Any advice will be greatly appreciated.
Thanks.
Mimosa

Hi Sebastian,
Thanks very much for your response. I thought the same but need second opinion to respond back to my client.
Could you advise me exact parameters allowed me to automatically stop long running/nonsense queries?
Thanks for your help.
Mimosa
Edited by: Mimosa on May 7, 2008 3:53 PM

BIA Sizing Information

Hi,
We are planning on buying BIA and here are the few questions asked by the Vendor for BIA Sizing - to get an estimate of how much memory and blades are required.
I am not clear on the few questions, can any clarify this...
*Initial Size (rows)*
Does this mean Init of Delta ? If its Init, then what should be included for the Cubes that have full loads
If its Init, then for the previous year cubes e.g. 2003 - the data could have been compressed then how to get that record count.
What if the Init was performed that was pulling quite a few years of data.
*Size of Delta Uploads (Rows)*
What if its a Full Load
What if the data is loaded from various source systems (R3, Flat File etc) and some are Full and some are Delta.
Do we need to provide the Delta Uploads even for the previous year cubes - in which NO delta will be done further.
*How will Distinct values change (e.g. 4 times more than today ?)*
What does this statement mean ?
Thanks, I will assign the points.

Hello Sesh,
first of all, sizing based on "t-shirt sizes" is not valid any more. We have better methods available for a long time already.
Memory Sizing
To ensure good performance with the BI Accelerator, the data of the BIA index must fit to an InfoCube in the main memory of the BIA hardware. The sizing approach differs slightly based on the following scenarios:
Scenario 1: BI System already exists and contains data to be indexed
For scenario 1 the memory requirement is estimated based on the existing data in the BI system. Since the memory consumption of BIA indexes is by far smaller than the memory consumption of the InfoCube on the database, SAP provides a program that analyzes your InfoCubes and estimates the memory requirement for the corresponding BIA indexes.
Note: You can already run the program in SAP BW Releases 3.0, 3.1 and 3.5. This means that you can carry out the sizing before you upgrade to SAP NetWeaver BI 7.0.
Install program ZZ_BIAMEMCONSUMPTION_BW3X in your BI system (see SAP note <a href="http://service.sap.com/sap/support/notes/917803">917803</a>) and run it for the selected InfoCubes. The output of the program contains the estimated memory per InfoCube and the estimated memory for the corresponding master data.
Scenario 2: BI System does not exist yet or no data has been loaded
For scenario 2 the memory requirement can be estimated based on some data points about the InfoCube data model. Number of dimensions, number of key figures, and number of records are used to estimate the size of the corresponding BIA index.
The number of SAPS for the BIA appliance can be estimated based on the type and number of users and certain assumptions about query complexity and usage frequency. This part of the sizing is done for the peak hour with the heaviest system usage.
Estimate the number of dimensions, number of key figures and number of records per InfoCube. The estimated memory per InfoCube and master data is determined by a complex formula. There's a spreadsheet available that contains this complex calculation (see SAP note <a href="http://service.sap.com/sap/support/notes/1010267">1010267</a>).
CPU Sizing
The requirement for number of SAPS is determined by the number and complexity of query requests that have to be processed by the BIA appliance during the peak hour. We assume that 195 SAPS are required for 100 query requests in an hour.
The number of query request can be estimated using the following formula:
Number of query requests = Number of Information Consumers * 1 + Number of Executive Users * 11 + Number of Power Users * 33
SAPS = Number of query request / 100 * 195
SAPS is a hardware independent unit to describe the CPU-related performance characteristic of a given hardware configuration. For more information on SAPS and their equivalent in hardware performance, see http://www.sap.com/benchmark > SAPS.
Regards,
Marc Bernard
SAP NetWeaver RIG
PS: Sizing has to be adjusted once new InfoCubes are supposed to be indexed by BIA that havent been taken into account for the original sizing.

Report avaibility is impacted due to BIA Indices

Hi BIA experts,
We are going to implement BIA for our BW reports. However, from the information that I gathered, I have a concern in here.
As I know before, when BIA Indices are not available, the report avaibility will not get impacted because it will automatically getting the data from DB (InfoCubes). Unfortunately, this is not the case in real life.
From the information that I gathered from people whom did the BIA implementation before.
When you have started using BIA Indices, when BIA Indices are not available, the reports which are having BIA indices will not be available until we fix the BIA Indices, all or nothing.
I heard there is an option to have DB/Aggregate fallback setting. When BIA indices are not available, DB/Aggregate fallback occured and the report will use aggregate/DB. This DB/aggregate fallback can be automatic or manual turned on.
My questions are:
1. What are the disadvantages using DB/Aggregate fallback setting? Will it impact BIA performance?
2. Except DB/Aggregate fallback, are there any other options to prevent report not available when BIA indices not available? eg. BIA index back-up (if there are such things).
Thank you very much for your response. Really appreciate.
Regards,
Mona

Hi Mona,
When BIA becomes unavailable, the BI system creates an entry(timestamp) to the table called RSDDTREXHPAFAIL. If you have email address(es) entered to the table called RSDDTREXEMAIL, then it sends an alert to those adresses. When this happens for the next thirty minutes all queries are diverted to BI ( or until you removed the entry from RSDDTREXHPAFAIL). This usually works when some servers become unavailable, and there is no disadvantage(other then not able to use BIA)
The problem here is sometime the system/server(s) hang, when this happens the BI system doesn't know the problem therefore doesn't create an entry. Then the users start receiving BIA errors. To fix this you go to RSDDBIAMON2 to switch the cubes off, yet if the blades are not available this transaction hangs sometime too. So we created adhoc programs to turn the cubes on/off + exclude/include users from BIA + include/exclude queries from BIA + and to create!/delete entries from the RSDDTREXHPAFAIL table.
Cheers
Tansu

How the data is fetched from the cube for reporting - with and without BIA

hi all,
I need to understand the below scenario:(as to how the data is fetched from the cube for reporting)
I have a query, on a multiprovider connected to cubes say A and B. A is on BIA index, B is not. There are no aggregates created on both the cubes.
CASE 1: I have taken RSRT stats with BIA on, in aggregation layer it says
Basic InfoProvider     *****Table type      ***** Viewed at      ***** Records, Selected     *****Records, Transported
Cube A     ***** blank ***** 0.624305      ***** 8,087,502      ***** 2,011
Cube B     ***** E ***** 42.002653 ***** 1,669,126      ***** 6
Cube B     ***** F ***** 98.696442 ***** 2,426,006 ***** 6
CASE 2:I have taken the RSRT stats, disabling the BIA index, in aggregation layer it says:
Basic InfoProvider     *****Table Type     *****Viewed at     *****Records, Selected     *****Records, Transported
Cube B     *****E     *****46.620825     ****1,669,126****     6
Cube B     *****F     ****106.148337****     2,426,030*****     6
Cube A     *****E     *****61.939073     *****3,794,113     *****3,499
Cube A     *****F     ****90.721171****     4,293,420     *****5,584
now my question is why is here a huge difference in the number of records transported for cube A when compared to case 1. The input criteria for both the cases are the same and the result output is matching. There is no change in the number of records selected for cube A in both cases.It is 8,087,502 in both cases.
Can someone pls clarify on this difference in records being selected.

Hi,
yes, Vitaliy could be guess right. Please check if FEMS compression is enabled (note 1308274).
What you can do to get more details about the selection is to activate the execurtion plan SQL/BWA queries in data manager. You can also activate the trace functions for BWA in RSRT. So you need to know how both queries select its data.
Regards,
Jens

How to check the BW and BIA systems are up and working fine.

Hi friends,
how to check the BW and BIA systems are up and working fine. And also what is ICM, how to check whether it is working properly or not.
ANd what is name resolution and IP , how to check whether this is running fine.
regards, balu.

Hi,
you can use process monitoring or the alert function in the standalone TREXAdmin tool.
ICM:
http://help.sap.com/saphelp_nw04/helpdata/EN/0a/a7903febb15a7be10000000a11405a/content.htm
name resolution
http://en.wikipedia.org/wiki/Domain_Name_System
IP:
http://en.wikipedia.org/wiki/IP
Please also have a look at the transactions TREXADMIN and RSDDBIAMON2.
I think it´s better if you use the forum search and google/wikipedia before you ask such questions.
Best Regards,
Jens

Few questions about "city explorer"

- I noticed it can tell me where the metro is, but could nokia maps + city explorer tell me what line should i take in the metro and which station i should go out?
- Does it all require data transfering while i'm walking, or can i donwload it all before, where i have wifi? (like nokia map-loader)
- same question about the guides...
Thank's!

You can find most of the answers here - or to links to the answers -
http://help.sap.com/businessobject/product_guides/boexplXI32/en/exp_acc_admin_en.pdf
I haven't used the Explorer accelerated for a long time, but as far as I remember in rsddstat i think you can identify the explorer usage from the user id, but not sure - try and see.
You can also run all your BIA indexes at the same time if you want to, the algorithm takes care of the table locks and switch to other index
Querysnap shot index is a SNAPSHOT of the data at the time of the index - it will never change unless you re-run the query and index it. Regular BWA indexes gets updated when you load data to underlying cubes and run BWA rollups/change runs. The querysnap shot data will not change. I like them if my query output miliions of lines - such as ZCUSTOMER. With query snapshot, you can put all of them to explorer for customer segmentation etc purposes
Cheers
Tansu

Questions about Trex migration

Hello experts,
I have some questions about Trex migration.
We have to migrate our Trex instances to another Hardware type (and operating system type).
Is it somehow possible to export the Trex database (indexes etc.) on the source Trex, and then import them on the new target Trex? I ask because i found another thread in here saying that this is only possible if the source and target OS is the same, and if the source and target Trex is the same version, I donu2019t know if this is true though?
Or is it easier to just install a new Trex, and then let it rebuild index etc.? And is this a feasible way to do it?
As you can see i don't know much about Trex, so any hints and good advice is appreciated.
Thanks in advance.
Regards,
Kenneth

Hi Kenneth,
there are different approachs for migartion TREX.
a) Install TREX on a new machine. And export the indexes and afterwards import the indexes. This is possible even you do not have the same TREX version. Because otherwise if you only want to upgrade your revision you just start the sapinst or install.sh at the same machine.
b) Intstall TREX on a new machine and re-index all.
But keep in mind TREX is not BIA. Even perhapts this is the same SAP software component.
So re-index could take about days.... not only 10 min. This depends on the type of objekts which should be indexed and of the volumne.
As well it is not recommanded in general to switch on delta. For BIA of cause but not in generel for TREX.
Best regards
Frank

BIA question...

Similar Messages

Maybe you are looking for