The hpc-ch Forum on Intra- and Inter-Site Networking has posted its Call for Participation. Hosted by the University of Zurich, the event will take place Thursday, May 19, 2016.
For HPC, data networking and data communication are critical elements in different contexts. Data is for example being moved between nodes during a computation, inside a cluster, data centre or between data centers to support a workflow or a distributed computation. With expected convergence of HPC and Data Analytics (sometimes referred as “Big Data”), requirements for data movement within and across data centers will additionally increase in coming years. Later is often referred as data federation. Today 100 gigabits of data per second can be transmitted over the network provided by SWITCH between Swiss research centers on up to 88 channels. Technologies used to transfer data inside a data center like Ethernet or InfiniBand develop very fast. Therefore, different transfer speeds across data centers have to be managed at the same time. New services that benefit from the increasing bandwidth within and across data centers are being introduced. An example is GridFTP, a high-performance, secure, reliable data transfer protocol optimized for high-bandwidth wide-area networks that can be used to transfer large amounts of data between sites.
This hpc-ch Forum will address the different aspects of intra- and inter-site networking including the different OSI layers from the physical up to the application layer, data federation concepts and use cases, and additional requirements such as security, authentication and authorization.
Key Questions
- What network technologies are you deploying inside your data center? What are typical networking architectures? How to interconnect the different segments?
- What kind of services supporting high throughput do we make available to outside users?
- How do we support cluster to cluster communication for computations (inter and intra cluster communication)?
- What applications and use cases currently use federated data? Are there standardized solutions that could be deployed?
