Resource Allocation Metrics Description
|Main||EGI.eu operations services||Support||Documentation||Tools||Activities||Performance||Technology||Catch-all Services||Resource Allocation||Security|
|EGI Resource Allocation menu:|
HTC Computing resources
Total computing time [HEPSPEC-hours]
By default time allocated in this metric is guaranteed. In case metric Opportunistic computing time is specified, the amount of guaranteed time allocated is lowered by the amount of opportunistic time.
Guaranteed computing time [HEPSPEC-hours]
The total calculation time intended for use by jobs in a grant on the basis of guarantees (computational resource consumption). Example: For two ongoing parallel jobs executing on ten slots for one day each the total calculation time is 2 * 10 * 24 = 480h
This metric limits the use of computing time available under warranty. After passing a limit set by this metric further use of the resources within this metric may not be possible.
The user obtains a guarantee that, provided appropriate jobs submissions, he will be provided with enough resources to use the entire allotted time. Appropriate submissions means commissioning jobs in a manner that allows the use of resources in a uniform manner throughout the duration of the grant.
Opportunistic computing time [HEPSPEC-hours]
The total calculation time intended for use by the jobs, not under warranty.
Metric limits the computing time used by jobs in situations where all resources granted within the metric GCT are gone. After using a limit set by this metric further use of the resources within this metric may not be possible.
Resource provider undertakes no obligation to ensure the availability of resources listed in this metric.
Max job duration [hours]
Maximum execution time for a single job. After this time, the job can be interrupted by site administrator.
Example: For grant with Max Job Duration metric equal to 72h job using four slots will be canceled after 72 real time hours (the number of slots used by the task does NOT affect the task time)
The user can expect (require) that a task with a duration of less than defined in this metric will not be interrupted.
Min local storage [GB]
The minimum scratch space used in the calculations for each slot involved in job execution.
The User can expect that the amount of scratch space specified in this metric will be the minimum available for each slot allocated for the job. The usage of more scratch space than specified may not be possible.
Min physical memory per core [GB]
The minimum RAM that can take a user application (job) on a single slot. The user can expect that if a job submitted uses no more RAM per slot than declared in the metric, it will be executed.
HTC Storage resources
Total storage capacity [GB]
By default capacity allocated in this metric is guaranteed. In case metric Opportunistic storage capacity is specified, the amount of guaranteed storage capacity is lowered by the amount of opportunistic capacity.
Guaranteed storage capacity [GB]
The target limit of capacity intended to long-term storage of data, on the basis of guarantees.
Opportunistic storage capacity [GB]
Additional non-guaranteed amount of disk space allocated for long term storage of data, increasing the limit for resource usage.
Cloud Computing resources
Total number of virtual cores
Total amount of computing capacity available for allocation measured in virtual cores. EGI virtual core is equal to EC2 virtual core (http://aws.amazon.com/ec2/).
Virtual Machines (maximum)
Maximum number of each type of resource template that may be hosted within the service. This may take into account the physical characteristics of facilities. There can be 4 types (flavours) of Virtual Machines offered by a site. Characteristic for each flavour:
'Small Virtual Machines'
- Number of virtual cores: < 2 virtual cores
- RAM [GB]: < 2 GB
- Scratch/ephemeral storage [GB]: < 20 GB
Medium Virtual Machines'
- Number of virtual cores: < 4 virtual cores
- RAM [GB] < 4 GB
- Scratch/ephemeral storage [GB] < 40 GB
Large Virtual Machines
- Number of virtual cores: < 8 virtual cores
- RAM [GB] < 32 GB
- Scratch/ephemeral storage [GB] < 80 GB
Extra Large Virtual Machines:
- Number of virtual cores: >= 8 virtual cores (OR)
- RAM [GB]: >= 32 GB (OR)
- Scratch/ephemeral storage [GB]: >= 80 GB
Allowed level of oversubscription
What is the state level of over provisioning that site is utilising in CPU and possibly memory. Range for the metric: 0% - 50%.
Total available storage volume made available to the user community, this should also note any over provisioning that is made of storage capability
High availability levels
Facilities or capabilities to provide high availability quality of service to user communities
Does the underlying facility have multiple network routes of entry (another form but more specific of high availability).