Site trail: Home ⇒ Service Level Agreements ⇒ JANET ⇒ 1995-96
Annex C
Notes on the Definition and Measurement of Service Levels
Contents
C1. Service Level Parameters
C2. Maintenance Time
C3. Measurement of Availability: Basic
Transmission Service
C4. Measurement of Availability: Other
Services
C5. End-to-End Service Levels
C6. Mean Time Between Failures
C7. Loss of Information
C1. Service Level Parameters
The criteria for choosing Service Level parameters are that they should be simple, easily measured, and meaningful to the users (technical staff and end users) at client institutions. The set of parameters defined for each service should cover availability, performance, reliability and the consequences of failure.
It is assumed that, as a matter of good management practice, UKERNA will make its own more extensive measures of performance in order to control its operations. Such measures are not, however, relevant to the Memorandum of Understanding.
C2. Maintenance Time
All the operational services shall be scheduled to be available 365 days a year and 24 hours a day, apart from planned maintenance times, which shall be published at least two weeks in advance to both primary and secondary sites. As far as possible, maintenance times shall be synchronized and shall fall during the "at risk" period from 08.00 to 10.00 on Tuesdays. The total maintenance time taken for any service shall not be more than 0.5% of service time, averaged over the calendar year. Service availabilities shall be calculated on the basis of scheduled availability (i.e. the time when the service was actually available to users divided by the time when it was scheduled to be available).
C3. Measurement of Availability: Basic Transmission Service
For transmission mechanisms under its control, UKERNA records the accessibility of each client institution from its Network Control Centre. In this context, "accessibility" means that traffic is observed passing in both directions across the client institution's access line. It would be simple to equate network availability with accessibility but inaccurate because of other operational factors: the reason for inaccessibility may be that the client institution has disconnected its equipment; and it may not be possible to determine the exact status of an apparently defective access line because no staff are available at the client institution (for example, at night or at weekends).
Network Availability will therefore be determined for each client institution according to the following definitions and procedures:
(a) "Staffed Periods" are the periods from
08.00 to 17.00 on all weekdays which are not Public or Bank
Holidays observed in all the constituent parts of the UK.
(b) Network availability from a client
institution is the time that network access was available divided
by the time that it was scheduled to be available.
(c) Network availability is calculated for each
client institution each calendar month.
C4. Measurement of Availability: Other Services
The measurement of availability of services other than the Basic Transmission Service is more straightforward. Availability of the other services is specified as follows:
(a) A service is "available" if it can be can
accessed via the network and used correctly by at least one
client institution in the case of services operated directly by
UKERNA. If a service is operated by a client institution acting
on behalf of UKERNA, it is available if it can be accessed via
the network and used correctly from any other client
institution.
(b) Availability is the time that the service
was "available" (as defined in (a) above) divided by the time
when it was scheduled to be available.
(c) Availability shall be calculated for each
service each calendar month.
Note that the availability of many services as perceived by client institutions will be the combination of the availability defined in (a) - (c) above and the network availability defined in section C3 above. UKERNA will report these two availabilities separately.
C5. End-to-End Service Levels
The service level parameters used in Annex A, including network availability as defined above, are measured within UKERNA's "management domain", i.e. only for the set of equipment for which UKERNA is responsible. It is the responsibility of client institutions to ensure that the service seen by their end-users, which comprises the combined services of the local network, the UKERNA service, and possibly another network which supports a remote host computer, is of suitable overall quality. It can nevertheless happen that problems occur which have a visible effect on the service seen by end-users but which cannot be traced with certainty to any of the separate management domains of the users' client institution, UKERNA, or the remote client institution. The following procedures will be used to resolve such problems:
(a) Individual client institutions shall, at
their own choice and on their own initiative, set up procedures
for measuring end-to-end service levels.
(b) Measurement procedures operated by client
institutions shall make reasonable use of network capacity and
shall not interfere with the ability of UKERNA to meet its
service level obligations to other client institutions.
(c) If measurements by a client institution show
a perceived service level to be significantly and persistently
below the corresponding value measured by UKERNA within its own
management domain, the client institution may call on UKERNA for
assistance in investigating the discrepancy and UKERNA shall make
reasonable efforts to provide such assistance.
(d) Client institutions shall make reasonable
efforts to assure themselves that perceived problems are not
caused by equipment in their own management domain (or in the
management domain of a remote client institution) before calling
for assistance from UKERNA.
(e) When the cause of a discrepancy has been
identified, UKERNA and the client institution shall determine by
mutual agreement whether UKERNA's records of service level
achieved need to be corrected and, if so, what the corrections
should be.
C6. Mean Time Between Failures
For certain services, it is useful to define Mean Time Between Failures (MTBF) as one of the Service Level parameters. In such cases, a "failure" is defined as any interruption in the operation of a service which causes a user of the service to lose any connection which the service provides or supports and which is not recovered automatically. Interruptions which cause the loss of connections but which are followed by automatic recovery of all connections active at the time of the interruption are not counted as failures.
C7. Loss of Information
Several services operated by UKERNA accept data submitted by client institutions and then make it available to the same and to other client institutions. UKERNA takes responsibility for the safe keeping of such data once it has been accepted. All procedures by which client institutions submit new data or update previously submitted data shall include a step in which UKERNA acknowledges receipt. UKERNA's responsibility for safe keeping starts when this acknowledgement is issued.
