QoS capability

[eswdd]

Various requests merge into one:

• We are looking for ways to implement a mechanism for rejecting requests that are coming for non-essential flows when we detect that a service has an out of ordinary load/response time. Impl by having a fully automatic mechanism that based on some rules to take the decision of blocking calls or to limit the number of calls based on flow priorities and/or agreed SLA’s.

Thoughts on how to implement:

Have a list of monitors (or a status aggregator) with thresholds that cause a trigger (seperate to impact on monitoring status), and then based on that execute pieces of logic to process a request at the entry to the EV (and maybe to it's Q), to allow rejection of requests with a reason.

[eswdd]

Currently working in the QoS branch. Have added a capability for having ExecutionPreProcessor which will be executed both prior to enqueuing in EV queue and also post. Have added a QoSProcessor which takes a Monitor, a triggering Status and an ExecutionPreProcessor to execute when triggered. Then implementation of those requirements above become relatively simple.

TODO:

• Construct some examples to meet the reqs above
• New ServerFaultCodes?

[eswdd]

QoS branch has now been merged into master. The Monitor has been changed to a StatusSource to enable a wider range of use cases.

[eswdd]

Have broken out the client time sections to #55 as this is less urgent.

[eswdd]

Gist describing how to configure this: https://gist.github.com/eswdd/7683613

[eswdd]

Marking as done since the load average monitor is not critical for the framework, only for the usecase that drove the feature. Documentation will be covered in the general doco issue #41