A number of Tideway customers have express a requirement for Tideway Foundation to discover installed software. The primary driver for these requests is to better enable software license audits when the license entitlement is based on installed software.

I’d be interested to better understand if other organizations have this requirement too and this is specific to any particular software vendors.

It would certainly benefit us, especially if version information were available.  In addition to license and audit issues, it would assist with troubleshooting by seeing perhaps host is not running the latest version of the software.

No doubt that feature can be of value. With however will come certain caviats and thus how you implement it will be of great importance.

Some thoughts:

Windows:
You need to implement great controls here to filter out the “standard” OS level installed application that register themselves with the installer. No user really cares that they have 700 copies of Notepad even though Windows likes to treat it like and application. Second, you need to leverage the same technology in your collection methodologies that other key finders use and get the license keys from the registry. This is crucial to providing the functionality for true license audits. It is understood that not all app vendors put their keys in the same place so this might be hit or miss.

UNIX:
Same goes here for the filtering however a new issue arrives in that not all applications register with the package installer so again, multiple means to obtain the data will be necessary.

Might be interesting to solicit what are the apps people want to find installed or to create a template diven collection mechanism that users can import common templates for production applications or build their own.

Tom

Perl can be installed with or without the OS installer utilities, -and- it can be installed anywhere.  It can be hard to manage installed perl libs from machine to machine.  “Does our fail-over server have the same perl libs as our primary server?”.

We would definitely like to be able to trigger on the existence of a directory or file…  we could use this to say “if /some/long/path/perl exists, then create an SI for Perl.”

I don’t mind having to hard-code directory paths into my rules, and would be willing to say “if /some/path/perl OR /another/path/perl OR /different/path/perl…” 
We’d also want to be able to confirm the existence of /usr/bin/perl, find where it’s linked, and parse the output of perl -V.

it is also very importend for us too!

This subject has come up internally here latley as well.  As I understand it.. the legacy BMC FD/TD product provides some support for this functionality today.. but am not familiar with it myself.

Any FD/TD folks out there that can expand on this topic?

Dave - 21 April 2010 03:05 PM

This subject has come up internally here latley as well.  As I understand it.. the legacy BMC FD/TD product provides some support for this functionality today.. but am not familiar with it myself.

Any FD/TD folks out there that can expand on this topic?

What do you need to know? :)
Basically, FD/TD provides discovery capability of installed products (synchronized to BMC_Product when synchronized to BMC Atrium CMDB).
Basic attributes that is discovered for each installed product are:

Product Name
Manufacturer
Version

For windows, these are discovered from Win_32 Product WMI class and some registry keys (depending on the method being used).
For aix, ‘lslpp’ I believe is used (the flags of this command CAN be crucial as the lpp or bff package name doesn´t make much sense)
For linux ‘rpm -qa’ or similar (same thing about the flags in the commands here, the rpm package name doesn´t make much sense)
Both on aix and linux one flag provides a user friendly name of the product.
For OpenVMS ‘product show product’ is used. (Known issue in this command is that its result only shows products that are installed in a certain way (can´t remember the name of it, but there is a thread about it on BMC Communities started by me about a year ago) The other methods that can be used for installing products on OpenVMS do not make the product show up with the “product show product” command. It is only available from a history file.

I am not 100% sure if FD/TD created the “ProductKey” (that are used when normalizing the data against the Product Catalogue) or if the ProductKey is generated in Atrium CMDB, but I guess it is generated in FD/TD.

Since FD/TD also stores a local copy of the Meta CIM model (CDM) and a local copy of the Product Catalogue from Atrium CMDB, FD/TD also normalize the product data against the Product Catalogue before inserting the data in the FD/TD datastore. This normalization includes, normalizing the Product Name, Version, Manufacturer and also Category, Type, Item (Tier 1, Tier 2, Tier 3) for the Product Catalogue. This last feature about Product Catalogue is also handled in the Normalization Engine in Atrium CMDB.

ok.. so “installed software” is primarily driven from the OS/platforms “installed packages” function.. whatever that may be.

Would I be correct in assuming that software installed via other means.. binary copies.. tarballs… whatever… and not officially “installed” through the platform installer is not covered?

Correct, my experience with FD/TD tells me that it is not covered and “installed packages” in ADDM 8 can be compared to what FD/TD provides.

Additional note, on Windows platforms FD/TD also discovers the Patches installed (BMC_Patch).

much appreciated.  Sounds like the only real gap at this point.. is that addm doesn’t model “Software Instances” related to Hosts from the current list of ‘installed packages’.

Dave - 21 April 2010 03:51 PM

Would I be correct in assuming that software installed via other means.. binary copies.. tarballs… whatever… and not officially “installed” through the platform installer is not covered?

This is by the way a really interesting question, as the question comes up: What is a “Installed Product”?
In my environment, I have to discover a “Installed Product”, which in fact is only 2 *.conf files in a directory, providing a “proxy” service and redirect incoming transactions. In my world, this would not be called a “Installed Product” or a “Running Software / SI”... I don´t know what to call it :)
If increasing the “scope” of Installed Products, then a statement clearifying what is considered an Installed Product would be needed.

Dave - 21 April 2010 04:46 PM

much appreciated.  Sounds like the only real gap at this point.. is that addm does model “Software Instances” related to Hosts from the current list of ‘installed packages’.

You lost me here :)
Mainly because I still have some difficulties learning the ADDM way of calling things, like Software Instance. I´m constantly trying to compare on or the other to a CI Class in the CDM.

I’ve never used FD/TD or Atrium before.. so I have no idea how the data gets mapped to into the CDM classes myself.

I know that when you do a search for “Software Instances” within ADDM… this is the list of software that has been discovered… but 99% of this is based on the presence of a running process on the host.. and not the presence of a file and/or registry key. 

Simply having one file… or a registry key present.. doesn’t really indicate that the full software package is available and functional.  If all the libraries and or conf files have been deleted.. but the binary executable is still there.. does that really count as ‘installed’?

its fairly easy to argue both sides of the ‘installed software’ predicament.  Vendors want their $$ for every cpu/core of every host the product is “installed on”.... but if its not really being used.. or the install is not 100% complete… should that count as a license?  Personally.. I tend to take the old tideway stance.. that if its not running.. it doesn’t exist.. and there is not actually a “software instance” on any given host.

Then again, within that is the fact that you need to discover the software when it is running… From a Atirum CMDB perspective I think that ADDM is somewhat limited to what can be discovered as everything discovered through a Pattern is synchronized to BMC_SoftwareServer (there are a few exceptions, I know).
As the Pattern Language is very powerful, (basically anything can be discovered and modeled), I think (personally) that each pattern created (custom created or TKU) should have the ability to set the CI Class to which it should be syncrhonized to in Atrium CMDB. Unfortunately there is no feature in Atrium CMDB to move a CI from one CI Class to another (since each CI Class has different attributes, depending on the CI type characteristics).

Take Microsoft Office as an example, where the licence type is per installed software… The chances are small that every user having MS Office installed, actually have it running when it is being discovered. Maybe a bad example but you get my point ;)

From this perspective, a “installed package” in ADDM, is a BMC_Product in Atrium CMDB, something we (can) count licensing on in Asset Management.

yep.. I see your point.. if the software is stop/started on demand.. there is no garuntee you will catch it during a scan…. unless of course you scan every system every hour or so… but that poses an entirley new set of issues when you have ~50k devices to scan.

so… “installed files exist.. vs running process actively seen during discovery”.  with neither answer being the best.

We can run a query for installed packages which roughly translates to installed software with some manual extrapolation

example query.