A research project to develop next generation solutions and products based on PCI Express over cable
PCI Express (PCIe) is today the most common technology to connect external devices such as graphics processing units, disk controllers and network adapters to a single computer.
PCIe over cable is also gaining momentum as a high-speed host-to-host network. One of the current challenges with PCIe is that there are no services that are able to integrate separate host to host communication and flexible transparent device access into a single hardware infrastructure. Existing PCI Express hardware can support this, but the required software does not exist - yet.
Devices connected to a single PCIe root complex do not support services such as hot-add, device migration and remote access. Devices can be addressed directly with the PCIe Non Transparent Bridging addressing technique, however, the required software infrastructure and OS interface do not exist. To transfer data between PCIe devices on different hosts today, the traditional networking services have to be used, adding unnecessary latency and overhead.
The goal with this project is to develop a new framework for operating system and virtual machines that will enable remote discover, addressing, access and use of standard PCIe devices.
The framework will enable standard PCIe devices to be re-allocated and shared by computer nodes in the PCIe network with no or minimum changes to the applications and device drivers.
On top of this framework we will develop services to validate and demonstrate the framework such as a fast cluster file system, legacy device driver access to IO devices on a remote node, and clustering of accelerator cards such as the Intel Xeon Phi coprocessor and Nvidia Tesla graphics processing units.
Dolphin Interconnect Solutions and Simula are the main research partners in the project. We are currently looking for early adopters bringing real use cases.
The project started in February 2014 and will last for 4 years.
The project is funded by Dolphin but receives significant contribution through the Norwegian Research Council BIA programme.
This page will be updated with project news and results.