Fri Apr 8 06:47:17 PDT 2016
Control Architecture: Change management: How are changes to information technology managed?
Option 1: Just let change happen and hope you can handle it.
Option 2: Make backups before changes in case you have to revert.
Option 3: Use change control prior to changes but move forward and count on your expertise to make it work.
Option 4: Use sound change control with full reversion capabilities.
Option 5: Recertify systems in context whenever significant changes are made.
Option 6: System-specific change controls should be used.
Just let change happen and hope you can handle it.
Fast and loose is often the best approach for smaller businesses or
for portions of businesses where risks are low. The cost of sound
change control is often on the order of twice the cost of not having
it, so low risk and low cost make sense together. But for medium and
high risk systems, change control is absolutely required.
Make backups before changes in case you have to revert.
Backups are used to provide a modicum of recoverability, and
changes are made with the knowledge that reversion is, at least
theoretically, possible. It is prudent to also have a tested recovery
procedure and to verify that thew backups are a workable solution for
recovery in desired time frames.
| || Research and development separated from change control separated from production, testing at each step |
Use change control prior to changes but move forward and count on your expertise to make it work.
In forward-only environments, such as financial transaction
systems, many enterprises choose to never go back once a substantial
change is made. This means that they accept the risk of failure and
save the cost of full reversion. It places additional pressure on
testing and process to get it right, and means that experts have to be
available to deal with the issues if and when they arise, and in real
Use sound change control with full reversion capabilities.
Sound change control implies:
- A system for requesting, specifying, implementing, and testing changes,
- A method for tracking and backing out of changes,
- Separation of duties between research and development, testing, change control, and operations,
- Databases that track these different elements of the process,
- Approval processes and work flows to assure operational execution,
- Integration of changes into the detection and response process to prevent false positives and potentially harmful responses,
- Notification of audit so they can adapt their auditing to meet the new requirements,
- Updated documentation to reflect operational changes and user changes,
- Training to adapt the people to the changes,
- HR and legal approval of changes impacting those areas, and
- Policies, standards, and procedures must be followed along the way.
Recertify systems in context whenever significant changes are
In high-consequence systems, particularly where process
failures can cause harm that is not rapidly and automatically detected
and readily mitigated, significant changes (i.e., anything other than
well-tested variations in parameters) should result in a system
recertification in context. This includes both a thorough retest of
the systems undergoing changes equivalent to the acceptance tests
performed initially, regression testing of all known and resolved
subsequent issues, and tests equivalent to acceptance and regression
tests of interoperation to assure that changes don;t have adverse
effects on related systems.
System-specific change controls should be used.
Copyright(c) Fred Cohen, 1988-2015 - All Rights Reserved
For systems that directly contact and control physical
systems, changes are very closely related to the specifics of the
physical system under control. As such, issues like stability and the
ability to maintain control over the process are central to change
management. As such, these are engineering decisions that rely on
engineering calculations and analysis. Such changes cannot be made in
bulk or based on a generic patch management approach, and cannot be
tested with common system testing methods. Plant shutdown is often
required for such changes, and simulations may also used to test some
classes of changes.