Why numbers, not wording
Maintenance agreements often contain phrases such as "we will make every effort" or "within a reasonable time". For an IT manager responsible for business continuity, such clauses give no basis for enforcement. The first good practice is therefore that every parameter has a numeric value, a unit and a reference period.
This applies to availability, service windows, backups and response times alike. Below we describe each parameter using the example of the InnerWeb maintenance agreement.
Availability: 99.5% in a calendar month
The InnerWeb maintenance agreement sets availability at 99.5% per calendar month for the server part of the instance. The calculation is simple: a 30-day month has 720 hours, so 0.5% is 3.6 hours, and in a 31-day month, with 744 hours, it is about 3.7 hours of unavailability.
When assessing such a parameter the IT manager should ask two things: whether the service window is included in the calculation and how availability is measured. Record in the agreement which outages are excluded and who measures them. The measured object must be the instance server, not the client's devices.
Permitted unavailability at different SLA levels
Response and repair times
In the event of a strategic failure, the response takes 12 hours and the repair 8 hours from taking the case on. It is worth recording the definition of a strategic failure: the IT manager may propose, for example, a situation in which the system does not work for most users or entries and permits cannot be issued.
Note that the repair time counts from taking the case on, not from the report. In the worst case the total time is 12 + 8 = 20 hours from the report, provided the response uses its full limit. If the plant's process requires a shorter period, record it as a separate requirement with a separate fee, or prepare an emergency procedure on the plant side.
Backups: frequency and retention
Backups are daily with 14-day retention. Daily frequency means that in the worst case, with a failure just before the next backup, data from one day can be lost. A 14-day retention means the state of the system from two weeks earlier can be restored, which is enough to recover data after a user error noticed late.
The IT manager should compare these parameters with their own requirements: is the loss of up to 24 hours of data acceptable for the register of entries, training and permits? If not, an additional procedure must be set, for example an event log kept on the plant side.
What to add to the backup clause
Record: frequency, retention, storage location, rules for restore testing and restore time after a report. A restore test matters more than making the backup itself, because a backup that cannot be restored is not a backup.
Worst-case time from a strategic failure report to repair
Service windows: up to 5 hours a month with notice
Service windows last up to 5 hours a month with notice of at least 5 working days. From the point of view of a production plant the time of day matters, not only the duration: a window in the middle of a shift at the main gate will block entries, while the same window at night from Sunday to Monday will go unnoticed.
Good practice is a clause stating that windows are planned outside peak entry hours and agreed with the IT manager, and that the notice goes to a named person, not to a general mailbox. Five working days gives time to warn the coordinators of external companies and security.
What the fee covers and what is charged extra
The maintenance fee includes: a dedicated instance, licences, the training module without a limit, backups, updates and security fixes, and support without billing of working hours. It follows that the IT manager does not have to negotiate backups or security fixes separately.
Requirements beyond this scope, for example a shorter retention or an additional test environment, should be priced separately and recorded as options. A clause on options makes later expansion easier for procurement without a new negotiation.
Checklist
Video
Frequently asked questions
How many hours is 99.5% availability?
In a 30-day month it is about 3.6 hours of permitted unavailability, and in a 31-day month about 3.7 hours. The result depends on whether service windows are included in the calculation.
How long are backups kept?
In InnerWeb offers backups are daily with 14-day retention. This allows the state of the system from two weeks earlier to be restored.
How much notice is given for service windows?
At least 5 working days. The total scale of windows is up to 5 hours a month.
Let's write the parameters into your agreement
We will go through availability, backups and service windows with your IT department before you sign the agreement. Book a meeting with the InnerWeb team.
Let's talk