Service Level Agreement - Science topic

very good question

I'm also interested in this.

This is a sample.

We referenced some standards in the enclosed papers - but note that our focus is security SLAs.

Hi Gabriel

The answer is: it depends.

Most universities have an Information Technology department responsible for provisioning all IT systems for the university. A cloud system is no different to a computer lab, or computing facilities provided for staff, or a high performance cluster (HPC) setup. In each case, the IT department will provide details of the standards of service (and any inter-departmental charging rates) to all users. For a cloud system, the provision of an SLA would provide a very prudent approach to indicate quality of service standards, especially where inter-departmental charges will take place.

Of course, some may take a less formal approach to service provisioning, and in this case, an SLA is less likely to be provided, often with a proviso that there are no guarantees on offer with the service. But there still needs to be some policies or procedures in place to ensure systems are not abused. This approach will often prevail in a Computing Science department where such systems are used for testing by CS students, and these systems are generally not available to mainstream users. However, for delivery of service to most other users, the provision of an SLA can be used as a good means of explaining the level of service that is on offer, and can be used to measure performance against these goals.

Regards

Bob

See my papers on SLA - you will find examples of datasets.

There are many software tools to monitor Cloud Computing performance for SLA purpose, some of them include in the following links:

Some of them do allow you to perform a Proof of Concept (PoC) for free evaluation.

Suggesting to consider the following before selecting & implementing a monitoring software tool:

1) what cloud workload performance needs to monitor for SLA generation - e.g. IaaS compute, memory, storage, I/O throughput, PaaS & SaaS software application types of SLA (availability, utilization, performance etc).

2) whether the software tool can monitor cloud computing performance based on multi-tenancy (if you are a public cloud service provider) or various-business units (even in a private cloud environment).

3) whether to acquire a monitoring software tool as 1-off purchase (capex), pay-per-use (opex) or cloud based SaaS etc (if via cloud based SaaS, need to clarify clearly the data & apps ownership when contract is terminated & location where the collected data is stored e.g. some companies / countries prohibit monitored data store outside the country).

4) how the SLA report being generated e.g. canned report, customizable report, required API integration with SLA Management Software / module from ITIL / ITSM Service Desk systems, required integration with other module for penalty / rebate credits calculation etc.

5) whether the software tool needs agent planted in the monitored nodes or is agentless (agent-based software tool needs additional effort to deploy the agent automatically or manually but can collect more information for detailed SLA calculation)

6) ease & completeness of training provided, design (HA or without HA), implementation, post-implementation support e.g. only 24 x 7 remote call support or onsite support within certain response time committed etc.

I was trying to do same but could not get. However, some secondary cases are available if you google it.

Be sure to make it clear that the IP rights to the software ( I am assuming the software was developed by you) belong to you and should not be violated. 

In cloud sim aleray provided SLA class you can directly use in to your programme ...... and you can define your own SLA factors view this papers your query will be solved

(1)https://www.researchgate.net/profile/Gandhi_Bipinchandra?ev=prf_highl

(2) https://www.researchgate.net/profile/Gandhi_Bipinchandra/publication/265058807_IRA_Technique_for_Desktop-as-a-Service_in_Cloud_Environment/links/53fdcf060cf2364ccc0911b0?origin=publication_detail

(3)https://www.researchgate.net/profile/Gandhi_Bipinchandra/publication/265058562_Survey_on_Resource_Management_Strategies_for_Desktop-as-a-Service_in_Cloud_Environment/links/53fdc8f20cf2364ccc090710?origin=publication_detail

this link should throw some light on your question.

We want to respect two SLA attributes: data transfer time and job execution time.

I.e., if we want to make sure that the data transfer between two Grid sites occur within specified time and job completes its execution on particular Grid site within specified time, how is it possible?

You may also take a look on an approach based on the concept of trust:

"Trust Measurements Yeld Distributed Decision Support in Cloud

Computing", International Journal of Cyber-Security and Digital Forensics (IJCSDF) 1(2): 140-151. The Society of Digital Information and Wireless Communications, 2012 (ISSN: 2305-0012).

You can find the full text at:

In my opinion monitoring as a service makes sense and should be part of the basic cloud SLA. Application performance monitoring (APM) feeds into many IT-governance processes like capacity planning, application life cycle management and risk analisis , to name a few. The data recorded by monitoring tools are therefore very important from a management perspective.

APM is also focussed on user experience monitoring and user experience should be included in the SLA (Well it is included in our SLA's). There is also another aspect I feel that should be added to the monitoring service and that is security monitoring. There is a move to offer Data Security as a Service (DSaaS) but that is not what I am pointing to here. I am talking more about Intrusion detection monitoring.

I think it is meaningful...and important in the long run. Of course, the negotiation time, when the demand is time critical may be a factor. Coud environment has enough factors to negotiate with the per-use pricing model. One could indicate parameters on which you are flexible with associated range, and the cloud system could propose solutions with relevant price ranges, and so on. Trade of memory, cpu power, completion deadline, priority, etc against price, or even against each other.