QoS Workshop
WLCG QoS Workshop
The purpose of the meeting is to stimulate WLCG's exploration the of potential of QoS, allowing better matching of systems to workflows and reducing the costs of storage provision. We hope to identify themes common at the experiment or site level which could benefit from coordinated action.
An updated version of the White Paper will be produced based on this and other feedback.
More information about QoS and the WLCG WG is available at https://twiki.cern.ch/twiki/bin/view/LCG/QoS
Call for contributions to the workshop
We would like to invite all stakeholders to present contributions relevant to storage QoS and its potential for WLCG.
Contributions dealing with the following topics are welcome:
- Experiment experience and plans with QoS
- Experiments are also requested to present their response to the White Paper
- Storage provider features and roadmap
- See below for further details
- Site experience with QoS
- See below for further details
Remote participation will of course be possible.
Please contact the organisers, Oliver Keeble (CERN) and Paul Millar (DESY), to schedule a contribution.
Further Information On Contributions
Storage Systems
We welcome all contributions from storage system providers, especially covering QoS-related topics such as:
- Multi-replica files
- Erasure encoding
- Multiple-media operation (Tape, Fast SSD, etc)
- User accessible QoS (Transitions, Media migrations)
- Data Lifecycle (Automated QoS transitions)
- Multi-site operation
- Geo aware placement and replica management
- Caching or volatile modes of operation
Site Contributions
We are particularly interested in contributions from sites about projects they have undertaken or conclusions they have reached regarding storage provision and cost.
In order to clarify the scope, here is a non-exhaustive list of possible topics:
- Redundancy
- Use of erasure coding in grid services
- Experience with RAID and its sustainability
- Integration of new redundancy layers in existing systems (e.g. Ceph behind DPM)
- Media
- Incorporation of novel media types
- Consumer vs enterprise drives
- SMR disks
- Fast cache
- Purchasing strategy
- Server densification, reducing overheads
- Data lake
- Replica management and multi-site storage systems
- Networking and implications for origin storage cost
- Stack consolidation
- Cost management through converging multiple user communities on the same system
- Self-classification of novel QoS classes and suggestions of how they can be exploited by experiments
- e.g. proposal for use of post-warranty hardware or new media types
- requests for use of novel (to WLCG) systems such as S3, offers of test services