SpiNNaker Machine Partitioning and Allocation Server (
This server enables users to request SpiNNaker machines of various shapes and sizes. These requests are queued and allocated to available SpiNNaker machines, partitioning large SpiNNaker machines into smaller ones if possible. When faced with the numerous research problems of optimal packing and scheduling of allocations, this implementation uses the ‘simplest mechanism that could possibly work’. With this said, the server intends to be ‘production ready’ in the sense that it is intended to be robust against real end users and routine maintenance requirements.
Server Management and Configuration¶
Essential reading for system administrators. The following documentation describes how to configure and manage a SpiNNaker Partitioning Server instance.
The following documentation describes the network protocol which clients must use to communicate with the server.
Essential reading for developers and maintainers. The following documentation describes the internal architecture of the SpiNNaker Partitioning Server and should serve as a guide to those wishing to extend its functionality.
- Internals Guide
- Development Installation
- Server logic (
- Top-level scheduling, allocation and hardware control logic (
- Job scheduling and allocation logic (
- Machine resource allocation at the board-granularity (
- Machine resource allocation at the triad-granularity (
- Number of baords to rectangle conversion (
- Asynchronous BMP Controller (