Inventory using uCMDB vs Flexera agent

[flexeranoob]

Dear all,

We are considering the pros vs cons of capturing the inventory using Flexera agent vs using teh uCMDB data for gathering inventory.  Could someone let us know the differences, or point us to a suitable link. 

TIA

[kclausen]

@flexeranoob

HPE-UD/uCMDB has a serious design flaw.  While UD does a very good job of scanning and capturing raw data, its recognition library to "normalize" the raw data to determine what software applications is installed is very poor, compared to other tools such as the FlexNet Manager ARL, or Flexera Technopedia.

The real issue, however, is that under the philosophy that a CMDB should only have clean/normalized data, when the discovery data is processed and imported into uCMDB, only the normalized software data is imported into the CMDB.  Any raw software evidence that is NOT RECOGNIZED will be ignored and is not imported into the CMDB.

Therefore, if you use uCMDB as your source of inventory into FlexNet Manager, you will not have a complete view of your software estate, as you will be stuck with the limited normalization capabilities of HP-UD/uCMDB.  Your only software inventory available to be imported into FlexNet Manager is the software that is already normalized.  You will not have any view of the unrecognized software evidence that gets discarded and not imported into the uCMDB.

[flexeranoob]

@kclausen Thank you for a detailed answer. This would imply that if we use uCMDB, the whole objective of using FNMS might be defeated.  Is it possible to use the raw UD data somehow?  We have a complex network and I am trying to remove any new port openings/agent deployment that we might have to do, and rather use an existing source of data. 

[mfranz]

Hi,

That sounds incorrect to me. We have a larger customer running uCMDB and the imported evidence data is actually a mix of raw inventory and applications already recognized by uCMDB.

The real problems is this: Both, raw data and already normalized apps, are run through the FNMS ARL. While raw data is usually fine, already normalized data can be tricky. E.g. in case of a false positive, it will be additional effort tracking down who's responsible for the false recognition. Is it a problem in uCMDB, or in the FNMS ARL?

Best regards,

Markward

[kclausen]

@flexeranoob 

@mfranz 

Yes, apologies.  What I described is the default behavior of uCMDB.  As described in the FlexNet Manager HP-UD documentation, you can configure uCMDB so that the raw Installation Evidence is imported into the uCMDB, along with the normalized inventory (which is what Markward has described).  Note that raw File Evidence is not imported into the uCMDB.

https://docs.flexera.com/FlexNetManagerSuite2020R2/EN/InvAdapConn/index.html#adapters/HP-UD/tasks/HPUD-Configuring%20HP%20Universal%20Discovery.html

[flexeranoob]

Thank for this.  So would you say that import from uCMDB is not recommended as such? 

[mfranz]

Hi,

There's more to consider. If uCMDB is already deployed and actively managed, why roll our a 2nd solution to basically collect the same data? Deploying the FNMS agent and keeping it deployed, including error tracking etc. is a considerable amount of work.

If there's special use cases, like Oracle DB inventory, or some Microsoft Server products, I guess the FNMS Agent might be worth it. Or perhaps a hybrid approach, deploying the FNMS Agent to servers and import the CMDB data? Why not start with importing from uCMDB, assessing the data quality and then decide from there?

Best regards,

Markward