RBS are looking at redesigning our server allocation and wondered what others are doing in regards to weightings, consumption and fallout?

We're not looking for the "correct answer" just a discussion on best practices and other CT customers experiences on how you're allocating your costs.

Thanking the community in advance

Hi Julie,

I'm not sure if this will help, but at ZNA we map our servers to an application or sometimes even multiple applications.  We have a CMDB where our servers are mapped to the application(s) and feed the mapped server data into our Servers Master data.  Fallout does occur if the server is not mapped, but that happens outside of Apptio since we only pull in our mapped servers. There is a monthly process to try to catch and identify all servers prior to the Apptio monthly close process, but of course inevitably something will fall through the cracks. This just prompts us to create the relationship and map the server to the application in the CMDB.

We outsource our infrastructure, so the costs for our servers come from the vendor invoice.  The vendor is required to include the application level information for each server on the invoice.  Any mismatch is usually found when there is a dummy application id used or perhaps a typo.  This requires us to research the issue and correct it with the vendor.  

This process works fairly well and we don't have too much fall out.  We create a Server_APP Key in both the Servers Master Data and the Applications Master Data and use this to create the data relationship between the two tables. 

For Sept 2019 our fallout was 0.1% and due to an application id not in our Applications Master Data.  Once we added that application id, it was 0.  (For the record, we have some "special" application ids that need to be manually added to our list.  Thankfully these are few and far between).

Hope this helps.  

Thanks Sue, are you using any weighting on the server to app allocation?

Hi Julie,

Yes we use server weightings. As an outsourced shop, our vendor invoice, which we use as a data source for Apptio, includes the server down to the application id level, so we leverage that information to calculate our weightings by application in the model.  We also keep the weightings in our CMDB (we use Service Now).  Here is a screen shot as to how we go about calculating the weightings by server.  It may not make complete sense, but hopefully it will give you an idea of what we do.  Hope this helps.

Thanks Sue that's very helpful

Hi @Julie Batty, I have a couple of inventories that i upload each month, until we get our CMBD updated with all our server info.

As our servers are outsourced we have a complex matrix for our unit charge which we have built within one of the transforms to create an allocation.

Our middleware also uses a unit charge which is then allocated to the servers

Our additional CPU & RAM costs use the unit charge which are assigned to servers.

Our storage is mapped to Servers that we can match to a server.

All our servers have an application name

As the Inventories are manually update i have lots of reports which catch a errors such as duplicates, incorrect apps names, shows if there are any storage, middleware not matching a server.

We also map the Environment, & have a DR (Disaster Recovery) column within the Inventory which again uses the unit charge for allocating costs.

Hope this helps

Hé @Lucas Wolters, maybe you can inform and advice @Julie Batty how we did the configuration in our CT model.

Hello Julie,

At what stage are you looking to modify the server allocation? Is it for allocating Server Costs to Applications or how the Server TCO is built up? We have created something quite granular when building up the Server TCO.

We're looking at how we allocate from servers to applications - currently we allocate server costs to the applications based on relationship (apps hosted on servers) and with a split weighting. The main reason for this is to see what is (& isn't working) for other CT customers with their infrastructure allocations 

Hi Julie,

Let me prescribe you the way we do it. This may seem a little complex but works very well for us and has helped reflect the reality close enough.

• Servers will either be running a database or hosting an application such as a web application or thick client application. 
• We first find the servers that host applications directly. In them, if more than 1 application is hosted we build a % of the server estate associated to that application purely by dividing it equally amongst all apps for eg. if a server hosts 3 Applications X, Y & Z....each app would bear a 33% of that server estate
• We then find out Servers that host Databases and which databases relate to which Applications and then connect the servers through to the Application using the similar logic in the previous step. 

We tried different logic such as utilization based or transaction based but figured that it leads to non-productive conversations

Thanks Suresh, that sounds like what we have currently. If you have the time would you mind elaborating on the utilisation/transaction logic leading to non-productive conversations? This is a route we're looking to go down. 

Sorry for the delay in responding @Julie Batty. 

So when we started considering the utilization and/or transactions based logic related to Apps, we figured that:

- it was getting too cumbersome to record the variation in utilization and/or transactions and let to more administrative effort. 

- it was becoming difficult to explain to the business on the variation in terms of utilization and/or transactions as those conversations were leading to reflect IT inefficiencies and Business would not be a partner on the variability and expected most of the things to remain stable.

thank you to everyone who has replied so far, this is very helpful