Example of multi-level CMDBf graph query

The CMDBf spec allows a multi-level graph query, but not to an undefined depth.  In other words, you must define a prototypical graph for the query; the spec does not currently support a prototype of variable depth.  Here's an example of a three-level graph query.

 Business service –Depends on-> Application System –Hosted by-> ComputerSystem

 I have a computer system; I want to know which application systems are hosted on it, and then which business services are dependent on those application systems.

<query xmlns="http://schema.cmdbf.org/datamodel"
 xmlns:xs="http://www.w3.org/2001/XMLSchema-instance"
 xmlns:cmdbf="http://schema.cmdbf.org/datamodel">

<itemTemplate id="i1">
  <recordTypeSelector namespace="http://van.org/itil_thingy" localName="BusinessService"/>
</itemTemplate>

<itemTemplate id="i2">
  <recordTypeSelector namespace="http://van.org/cim_thingy" localName="ApplicationSystem"/>
</itemTemplate>

<itemTemplate id="i3">
  <instanceIdSelector>
    <mdrId>myMdrId</mdrId>
    <localId>myComputerId</localId>
  </instanceIdSelector>
</itemTemplate>

<relationshipTemplate id="r1">
  <recordTypeSelector namespace="http://van.org/cim_thingy" localName="Dependency"/>
  <source ref="i1"/>
  <target ref="i2"/>
</relationshipTemplate>

<relationshipTemplate id="r2">
  <recordTypeSelector namespace="http://van.org/cim_thingy" localName="HostedDependency"/>
  <source ref="i2"/>
  <target ref="i3"/>
</relationshipTemplate>

</query>

The implementation could start with the most specific itemTemplate (i3) and traverse upward (to avoid processing very large intermediate result sets).  Therefore as long as I am willing to accept a fixed depth, and possibly create multiple queries (one for each possible depth), I can accomplish what I want to achieve.

The committee has considered adding semantics to support variable depth queries, but no decision has been made.

Technorati Profile