Snellius and Lisa, similarities and differences
Lisa and Snellius both are compute facilities, clusters on which compute jobs can be run that utilise one node, or multiple nodes in parallel in an orchestrated way. However, both systems are funded differently and consequently options and conditions for obtaining access to either or both of these systems may not be the same for all potential users. In addition, both compute facilities are definitely systems in their own right, with different node flavours and other resource features.
While both systems are optimised to match their specific resources, SURF strives to optimise commonality between the two systems from a user perspective. In our view, it should be easy for users to go from one system to another in either direction, or to have projects on both systems, without a steep learning curve.
Snellius
Snellius is the Dutch national supercomputer. Snellius is a general purpose capability system and is designed to be a well balanced system. If you need one or more of: many cores, large symmetric multi-processing nodes, high memory, a fast interconnect, a lot of work space on disk, or a fast I/O subsystem then Snellius is the machine of choice.
Snellius is accessible at the address:
snellius.surf.nl
For questions and requests please open a request at our ServiceDesk.
Information on getting access to Snellius can be found here.
Quick system overview
( see here for detailed hardware information)
# | Type of node | Characteristic |
---|---|---|
3 | interactive login | 16 cores/node |
504 | "thin" CPU | 128 cores/node 256 GiB memory/node |
72 | high memory "fat" CPU | 128 cores/node 1 TiB memory/node |
2 | very high memory CPU | 128 cores/node 4TiB memory/node |
2 | very high memory CPU | 128 cores/node 8TiB memory/node |
36 | GPU | 72 CPU cores/node 4 attached NVIDIA A100/node |
7 | Service | 16 CPU cores/node |
Interconnect | ||
Infiniband HDR100 (100Gbps), fat tree |
Lisa
The Lisa Cluster is meant for those in need of large computing capacities, but do not need the facilities of a real supercomputer. The system offers a large number of multi-core nodes.
Lisa is accessible at the address:
lisa.surfsara.nl
For questions and requests please open a request at our ServiceDesk.
Information on getting access to Lisa can be found here.
Quick system overview
(See here for detailed hardware information)
# | Type of node | Characteristic |
---|---|---|
2 | interactive login CPU | 16 cores/node |
1 | interactive login GPU | 12 cores/node 4 GeForce 1080Ti/node |
192 | "thin" CPU (Gold 6130) | 16 cores/node 96 GiB memory/node |
96 | "thin" CPU (Silver 4110) | 16 cores/node 96 GiB memory/node |
6 | "thin" CPU (gold_6230R) | 52 cores/node 384 GiB memory/node |
1 | high memory "fat" CPU | 48 cores/node 2 TiB memory/node |
23 | GPU GeForce 1080Ti | 12 CPU cores/node 4 GeForce 1080Ti/node |
2 | GPU Titan V | 12 CPU cores/node 4 Titan V/node |
29 | GPU Titan RTX | 24 CPU cores/node 4 Titan RTX/node |
Interconnect | ||
CPU nodes | 10 Gbit/s ethernet | |
CPU nodes (gold_6230R) | 2 x 25 Gbit/s ethernet | |
GPU nodes | 40 Gbit/s ethernet |
Common elements
Batch and reservation system
Snellius and Lisa use the same batch system - SLURM - though at any time the supported version may diverge. Snellius and Lisa use the same accounting and budgeting tools, that are a site-specific add-on to the tools provided by SLURM.
Software, applications, libraries, and tooling to build your custom applications
SURF strives to have a large number of scientific application software packages, libraries, and associated tooling available on both systems. The supported version(s) of any particular application or library may diverge between Snellius and Lisa. The software is built in such a way that it is optimised for the node flavours and node-interconnect of each system. This may imply that specific optimising features are not applicable to either system.
Documentation
The overviews and guides listed below are generic in that their contents applies to both systems - and parts of it possibly even to other HPC infrastructures elsewhere.