SNP Runtime

[former_member928177]

Hello Gurus,

I got problem while running the SNP OPT. as i have scheduled the run in background at night for 6 hours.But next morning found that run is till running, i have given time to complete it, But till after 35 hours has been completed the run is still active. Killed the job.

In settings for optmizer profile time mentioned for max. runtime is 360 min.

Why it was still running for this much time?

As mentioned time in opt. profile i.e. max time is 360 min. it should end the process but not WHY ?

Help will be appreciated..

Amol Chavan

[Former Member · Accepted solution]

HI

At which stage was the Optimiser on, 1, 2 or 3?

Stage 1 is the data load

Stage 2 is the optimisation calculation

Stage 3 is the results writing.

It is only stage 2 that is controlled by the maximum runtime setting.

[former_member928177 · Accepted solution]

Hello Ian,

Thank you for your reply.. Opti rin got stuck in Stage 2 only. Because the stage 1 has been completed but not Stage 2 . Is there any possibility that if you enter same location twice in location field.(Just for sake)? Got run once again it was successful but not giving the optimal solution.. How to trace it abt what went wrong?

Thanks,

Amol Chavan

[Former Member · Accepted solution]

Okay

first thing to check is in transaction /sapapo/opt11 and display the trace file. Does it have any specific messages in there (Look for SnoModule onwards).

In transaction /sapapo/perfmon select statistics: SNP Optimizer and double-click your last run. How many variables and constraints does it tell you your model has?

[Former Member · Accepted solution]

Hai Amol,

How you have come to a conculsion that system has not provided optimum solution?

Which version you are in? If you are in SCM 5.0, explanation tool would provide quite a handful of information, why the demand could not be fulfilled?

If you are in lower versions (SCM 4.0, 4.1 or below), you need to analyse through the input logs and output logs.

Regards,

Murali.J

[former_member928177 · Accepted solution]

Hello Murali,

As the solution provided gives maximium cost which is not the case always for the same demand. I am using 4.1 How should i correctly cheak the input o/p log..

Amol

[Former Member · Accepted solution]

Hai Amol,

If you can give me more details of your scenario, I will try to help you.

i.e. Details about how you have configured your network model, the different type of costs updated in your masters and details about optimizer profile.

Regards,

Murali.J

[former_member928177 · Accepted solution]

hello Murali and all gurus,

I tried once again in simulation version for the same run. But failed..

Settings made are : We used to run SNP for 9 plants. transportation lanes are maintained product specefic. abt optimization profile considering all capacity constraints, zero lmin ot size in weekely buckets, stop criteria 120 min with 9999999 improvements , no decomposition , first heuristic solution, objective of min. cost, dependent n distribution demand as pseudo hard

What could be the probable reason..

Please reply...

Amol Chavan

[Former Member · Accepted solution]

HI

Please can you tell me How many variables and constraints does it tell you your model has? This will explain how complex your model actually is.

In transaction /sapapo/perfmon select statistics: SNP Optimizer and double-click your last run.

[Former Member · Accepted solution]

Hai Amol,

Are you running with discretization or linear optimization? If you are running with discretization and experiencing run failures, you can choose linear optimization in the profile and try.

Are you in 32-bit optimizer hardware environment or 64-bit hardware? If you are in 32-bit environment, with Windows enterprise edition operating system, maximum processor memory possible to achieve is around 2.8 GB (at the most max. 3 GB)

You can see how much of the memory your optimizer run has hit by viewing in /n/sapapo/opt11 - trace file - parameter: "peak virtual bytes". Please let me know the reading you are observing here and response to the above parameters.

If you are hitting peak values in all of the above, then there are options such as going for decomposition functionality or splitting your optimizer runs to reduce the variable size and finally go for 64-bit environment with SAP consultation.

Hope this helps you and I will be able to further provide you information based on your inputs.

Regards,

Murali.J

[former_member928177 · Accepted solution]

Dear Murali,

One thing i want to share if i Run SNP for single FERT it gives results i.e. it is successful. Again i tried for the same for 360 min. Which got stuck for 2nd stage. We are using 64 bit hardware with sufficient processor memory.In opt11 it gives the details :

No. Of constraints : 1511180

Discrete variables : 54542

variables : 243021

peak_memory : 1523996

peak_virtual_memory : 1743308

real time :360

What is that will take decision for peak virtual memory.

Regards,

Amol chavan

[Former Member · Accepted solution]

Ok Amol

The first thing to note is that you are running the Discrete Optimiser, which in itself takes a lot longer than the Linear version.

What are you discretising on?

Is it more than one discrete item, if so then it will take a very long time to complete even a fairly simple model, though again 1.5million constraints is not a simple model.

[former_member928177 · Accepted solution]

Dear Ian,

It is not the issue that we are improving the system run to Discritization. Prior to this we used to run on the same settings that are made, which are giving the results as per requirement. But since we have mapped new plant into SCM results are failing For that we have one simulation Run for 2 FERTS which was successful( For theplants including newly added.) Got the result for new plant. But when we are doing for whole supply chain it is stucking in between..

Can you suggest the things to have a cheak...

Regards,

Amol chavan

[Former Member · Accepted solution]

Hi

The reason why I am asking about your discrete settings is that you have a large number of discrete variables there:

No. Of constraints : 1511180

Discrete variables : 54542

variables : 243021

You are hitting the maximum runtime, hence real_time:360 without a solution being found. I don't think it is getting stuck, I think it is having a problem defining a solution for that number of discrete variables. A run for two FERTS would have a lot less variables to consider and hence will complete successfully.

[former_member928177 · Accepted solution]

Dear Ian,

What could be the way to reduce the no. of variables. i am getting confused..

Since prior to this system is giving results on the same no. of variables.

Amol Chavan

[Former Member · Accepted solution]

The number of variables will depend on lots of things, the main ones for dicrete optimiser are:

Number of PPMs with minimum lot size requirements

Number of PPMs with discrete lot size requirements

Number of PPMs with fixed resource consumption

Number of production resources with discrete expansion

Number of discrete transportation methods

Number of transportable products with minimum lot size requirements

Number of transportable products with discrete lot size requirements

Average number of segments for piecewise-linear cost functions

For each of the above that you are using, you then multiply them by the numbers of buckets you are using in the planning horizon.

If you add in a new location this will affect some or all of the above variables as your new location will possibly have PPM/PDS's associated with it, extra transport lanes and products on the lanes. Possibly extra production resources etc etc.

Each item of master data added to a model increases the number of variables in the Optimiser and hence the complexity.

SAP recommend that no more than two factors are considered in a discrete optimisation run and preferably only one is used (e.g. discrete lot size).

Also, over what period do you need to have a discrete result? Can you reduce the discrete horizon and let the optimiser calculate the result in a linear manner for the rest of the horizon?

Can you split the optimisation run into smaller runs so that you can have multiple optimisation jobs running in parallel?

See OSS note <a href="https://websmp103.sap-ag.de/~form/handler?_APP=01100107900000000342&_EVENT=REDIR&_NNUM=454433&_NLANG=E">454433</a> for more details on how section 2 is affected by discretization.

Good Luck

Ian

[Former Member · Accepted solution]

hi Amol,

did u get the reason of ur run-time prob?..if not then plzz check ur opt-profile's 'extended settings' tab. and check if the ' time-based constrain' box is checked or not. If u r using an SNP optimization bound profile for the SNP optimization run, then ignore 'time-based constrain' .

best of luck.

Satyajit

[former_member928177 · Accepted solution]

Dear Ian,

Can you brief me about Transportation lanes. Is it necessary to maintain all products in TL from any logical location to plant, As we have maintained near by 190 lanes and most of them are maintained for all products which are maintained at different plant locations. e.g. TL is from Export(DC) to 1000 plant for FERTs if this lane being maintained with products that are not maintained in 1000 plant. Will it cause any effect while planning.

Further development is that at least run is getting failed with in stipulated time, since it is not finding any results...

Discrete lot size and its effect on planning run ?

Please ... Please ...help me out as i am having no production run since then..

Regards,

Amol Chavan

[Former Member · Accepted solution]

Hi

There are several ways you can maintain products on a transportation lane, the easiest setting is to activate the product-specific transportation lane setting with a single product procurement settings of "all products". This will mean that this lane is valid for all products between these two locations. This should not cause you any problems if the product does not exist in your 1000 plant as it will not then use this transportation lane in planning this product (you can see if it does in your optimisation log under the input table ET_ARCMAT, if no entry on the lane for your product then it has not been considered).

Another way, is to maintain the products individually on the transportation lane, this says to the optimiser that only the products I specify here can move on the lane (if all products can move on the lane then use the top option and not this one to save master data maintenance). If the product does not exist in both locations you will potentially have problems in the planning run but these should be captured as messages in the message log.

There is an option (user-specific) to check on entry that they exist in both locations. In the transaction (/sapapo/scc_tl1 click on settings -> settings then check the box "Existence of product at start and end locations" and it may be an idea to check also "Assignment of location products to model"). This will ensure that any future entries are validated. You can run the model consistency check to find those products that do not exist in your location (trans: /sapapo/conschk). Here you need to ensure that the above settings are checked in your Model Check profile.

Have you investigated the "message log" output from the optimiser logs to see if there are any indications here of irregular data or potential problems?