Providing IO distributed system
Technology Deep Dive
What is ExpEther ?
ExpEther is a technology that can extend the PCI Express bus beyond the confines of a computer chassis via Ethernet without any modification of existing hardware and software. Computer resources can be added to a standard Ethernet fabric as if adding such resources directly into the chassis to provide scale-up flexibility. ExpEther can build a new type of computing environment without localized physical constraints and is cost effective with the use of standard Ethernet equipment.
Broad-Scale Single Computer
ExpEther can build a new type of computing environment without any localized physical constraints. There is no limitation of distance between a host and I/O devices though it depends on allowable latency of applications. In an actual deployed system at Osaka University, the longest network path between the host and I/O devices was more than two kilo meters.
Full Compatibility with PCI Express Specification
The ExpEther Engine is designed to be seen as a PCI Express Switch from the CPU. The CPU is unaware of the Ethernet Fabric in between. The ExpEther Engine is fully compatible with PCI Express Specifications and has passed the PCI-SIG PCI Express 2.0 Switch's compliant test. It currently appears on the PCI Express 2.0 Integrators List.
System Configuration in Multiple Servers Environment
Each ExpEther Engine has a "Group ID" to connect a Host and I/O devices logically. This Group ID is assigned through a hardware rotary switch or by using Management software (ExpEther Manager). The ID range is from 1 to 4,000 and is used as a VLAN tag to isolate each PCI Express path.
If the ID of an ExpEther Engine is changed, the I/O devices under the ExpEther Engine are logically removed from the host by following the PCI Express Hot-Plug scheme.
More Technology Information
If you are interested in ExpEther technology, please contact us.
The ExpEther Consortium can provide more technical information.
Research Activity using ExpEther
by Hiroki Matsutani, Associate Professor, Dept. of Information and Computer Science, Keio University
|Virtual Reality using Remote GPU over 10GbE||Accelerating Apache Spark using Remote GPUs|
|Accelerating MongoDB using Remote GPUs|