The Ghost CI Cleanup
A financial services company has a CMDB with 120,000 configuration items (CIs). The CMDB manager notices that 35% of CIs have not been updated in over 18 months. Many of these are servers that were decommissioned but never marked as retired. The company uses discovery tools that run weekly, but the discovery tool only covers 60% of the network. The CMDB is used by the incident management team to assess impact, but they frequently find that the CI they need is missing or outdated. The CIO has asked for a plan to improve CMDB accuracy and reduce the number of stale CIs. The CMDB manager has a limited budget and can only implement one major improvement this quarter.
1. What is the most likely root cause of the stale CIs, and what single improvement should the CMDB manager prioritize to address it? Justify your answer.
Read the worked answer
The root cause is that discovery coverage is incomplete (only 60% of the network), so many CIs are not being automatically updated. Additionally, there is no process to retire CIs when servers are decommissioned. The single improvement should be to expand discovery coverage to the remaining 40% of the network, because automated discovery is the most reliable way to keep CI data current. While implementing a retirement process is also important, without full discovery, manual processes will still leave gaps. Expanding discovery will automatically update or flag CIs that are no longer responding, reducing stale records. The manager should also configure the discovery tool to mark CIs as 'retired' when they are not found for a certain period, but this is part of the same discovery improvement.
2. The incident management team complains that they cannot rely on the CMDB for impact analysis. How would you measure the success of the improvement in terms of CMDB data quality? Propose two specific metrics.
Read the worked answer
Two metrics: (1) Percentage of CIs updated by discovery in the last 30 days (target >90% for discovered CIs). (2) Percentage of CIs with a status of 'retired' that are still being discovered (target 0%). Additionally, track the number of incidents where the CMDB was used and found to have accurate CI information (e.g., via post-incident survey). Success would be a significant reduction in stale CIs and increased confidence from incident management.