Configuration Management:Layout 1 10/13/10 4:59 PM Page 107
POPULATING A CMDB: PROCESS DESIGN
is often one big bucket of different server and network repositories without a
centralised location for holding physical assets. Nevertheless, automation has its
issues too, such as tracking the installation of agents, making agents work over
the internet, and tracking virtualised services across physical servers.
Automated import of desktop hardware information can be easy enough, but
location by IP address or services used is difficult, and this may always be
necessary.
Integration of information about servers, switches, links, printers etc., held in
different files and tool formats can be an issue. One delegate said, ‘Our tool has to
get at this via export from our management system to Excel
®
, then by importing
Excel
®
CSV format, so relationships get lost.’
Identification – data type = Other
You need to identify software and software versions running on the server estate,
of course, but sometimes a client requires items like PDAs to be captured, which
can be difficult to maintain. Maintenance is an issue: you have to decide which CIs
to record data for, on a basis of, ‘if you want it, how are you going to maintain it?’.
Relationships can be tricky too, where (for example) a CI is dependent on multiple
CIs but not necessarily all at the same time, such as when this server or that
server must be up but both don’t need to be up at the same time.
Categorisation – Details
As usual, data quality and data accuracy were fundamental issues. One delegate
said, ‘We don’t have good attribute information or aren’t sure information is
correct.’ With outsourced infrastructure, where some items are controlled by third
parties, a lack of confidence in the quality of data is quite common. You need a
way to coordinate audits of all details both outsourced and in-house.
You need to be able to identify critical applications and decide on ownership and
governance, and use this to help to prioritise understanding and definition of a
CI and attribute level that will appeal to and work for both IT and the business.
You should anticipate possible conflict over system and service categorisation by
business area, function or personal opinion. As always, the CMS team needs good
people and negotiation skills.
Categorisation – Status
This can be a significant challenge. If a CI is owned by business users its status
may often be effectively unknown (this is an issue to address, not a desirable state
of affairs). You need to record the status of changes against each application as
well as knowing the status of assets such as servers, workstations and facilities.
The decommissioning process can also be a challenge, especially with automated
discovery: has a CI been decommissioned or has its host simply been switched off?
There is a tendency not to remove things from the CMDB in case they are
only temporarilly absent, although perhaps they should be recorded as ‘retired’
(but then, do you want to distinguish between ‘scrapped’, ‘disposed of/sold’ and
‘stolen’?). You need to think of decommissioning as a process and decide what
information is worth keeping and why.
107