Katana 是位于新南威尔士大学校园内的一个共享计算集群，旨在为处理非敏感数据的团队提供轻松访问计算资源的方式。它包含超过 6,000 个 CPU 核心、8 个 GPU 计算节点（V100 和 A100）以及 6Pb 的磁盘存储。Katana 提供了一个灵活的计算环境，用户可以在其中运行在台式机或笔记本电脑上不可能或不实用的作业。有关计算节点的完整详细信息（包括完整列表），请参阅下面的计算节点信息部分。
Katana 可以提供完美的培训或发展领域，然后再利用所获得的知识转移到更强大的系统，例如位于国家计算基础设施 (NCI) 的澳大利亚峰值 HPC 系统 Gadi。如果您不确定高性能计算是否适合您，或者有本地支持、节点混合、较长的工作运行时间和广泛的软件（特别关注生物科学），它也可能是一个完美的地点，它可能是一个完美的地点为您的研究管道。
About Katana
Katana is a shared computational cluster located on campus at UNSW that has been designed to provide easy access to computational resources for groups working with non-sensitive data. It contains over 6,000 CPU cores, 8 GPU compute nodes (V100 and A100), and 6Pb of disk storage. Katana provides a flexible compute environment where users can run jobs that wouldn't be possible or practical on their desktop or laptop. For full details of the compute nodes including a full list see the compute node information section below.
Katana is powerful on its own, but can be seen as a training or development base before migrating up to systems like Australia's peak HPC system Gadi, located at NCI. Research Technology Services also provide training, advice and support for new users or those uncertain if High Performance Computing is the right fit for their research needs.
System Configuration¶
RPM based Linux OSes. Rocky on the management plane and nodes
PBSPro version 19.1.3
Large global scratch at
/srv/scratch, local scratch at$TMPDIR12, 48, 100, 200 hour Walltime queues with prioritisation
Compute¶
Heterogenous hardware: Dell, Lenovo, Huawei.
Roughly 170 nodes
GPU Compute¶
The most popular use of these nodes is for Tensorflow.
Eight GPU capable nodes
Tesla V100-SXM2, 32GB
Nvidia A100, 40GB
Five are dedicated for the department that owns them
Three are general use for all researchers
You cannot use Tensorflow on the login nodes because they don't have GPUs. You will need to get access to a GPU node to do this.
如需申请使用Katana，请访问 https://research.unsw.edu.au/katana
