HPC Center user guides

_images/HPC.jpg

The use of the resources of TalTech HPC Centre requires an active Uni-ID account (an application form for non-employees/non-students can be found here). Further the user needs to be added to the HPC-USERS group, please ask hpcsupport@taltech.ee to activate HPC access from your TalTech e-mail and provide your UniID (six letters taken from the user’s full name). In the case of using licensed programs, the user must also be added to the appropriate group. Here can be found more about available programs and licenses.

TalTech HPC Centre includes cluster, cloud and also is responsible for providing access to resources of LUMI supercomputer.

The cloud provides user ability to create virtual machines where the user has full admin rights and can install all the necessary software by her/himself. VMs can be connected from outside and can be used for providing web services. Accessible through the ETAIS website: https://etais.ee/using/.

The cluster has a Linux operating system (based on CentOS; Debian or Ubuntu on special purpose nodes) and uses SLURM as a batch scheduler and resource manager. Linux is the dominating operating system used for scientific computing and of now is the only operating system present in the Top500 list (a list of the 500 most powerful computers in the world). Linux command-line knowledge is essential for using the cluster. Resources on learning Linux can be found in our guide, including introductory lectures in Moodle. However, some graphical interface is available for data visualisation, copy and transfer.

LUMI supercomputer is the fastest supercomputer in Europe, the fifth fastest globally and the seventh greenest supercomputer on the planet. Specification of LUMI can be found here.





Hardware Specification


TalTech ETAIS Cloud: 5-node OpenStack cloud
  • 5 compute (nova) nodes with 768 GB of RAM and 80 threads each

  • 65 TB CephFS storage (net capacity)

  • accessible through the ETAIS website: https://etais.ee/using/

TalTech cluster base (base.hpc.taltech.ee):
  • SLURM v23 scheduler, a live load diagram

  • 1.5 PB storage, with a 0.5 TB/user quota on $HOME and 2 TB/user quota on SMBHOME

  • 32 green nodes, 2 x Intel Xeon Gold 6148 20C 2.40 GHz (40 cores, 80 threads per node), 96 GB DDR4-2666 R ECC RAM (green[1-32]), 25 Gbit Ethernet, 18 of these FDR InfiniBand (green-ib partition)

  • 1 mem1tb large memory node, 1 TB RAM, 4x Intel Xeon CPU E5-4640 (together 32 cores, 64 threads)

  • 2 ada GPU nodes, 2xNvidia L40/48GB, 2x 32core AMD EPYC 9354 Zen4 (together 64 cores, 128 threads), 1.5 TB RAM**

  • amp GPU nodes (specific guide for amp and amp1):

  • amp: 8xNvidia A100/40GB, 2x 64core AMD EPYC 7742 Zen (together 128 cores, 256 threads), 1 TB RAM;

  • amp2: 8xNvidia A100/80GB, 2x 64core AMD EPYC 7713 zen3 (together 128 cores, 256 threads), 2 TB RAM

  • Visualization node viz (accessible within University network and FortiVPN, guide for viz): 2xNvidia Tesla K20Xm graphic cards (on displays :0.0 and :0.1), CPU Intel(R) Xeon(R) CPU E5-2630L v2@2.40GHz (24 threads), 64 GB RAM, HDD 2 TB storage.




Billing


Virtual server hosting

What

Unit

TalTech internal

External

CPU

CPU*hour

0.002 EUR

0.003 EUR

Memory

RAM*hour

0.001 EUR

0.0013 EUR

Storage

TB*year

20 EUR

80 EUR

TalTech cluster

What

Unit

TalTech internal

External

CPUcore & < 6 GB RAM

CPUcore*hour

0.006 EUR

0.012 EUR

CPUcore & > 6 GB RAM

6 GB RAM*hour

0.006 EUR

0.012 EUR

GPU

GPU*hour

0.20 EUR

0.50 EUR

Storage

1 TB*Year

20 EUR

80 EUR

More details how to calculate computational costs for TalTech cluster can be found in Monitoring resources part of Quickstart page .

LUMI cluster for users from Estonia

What

Unit

Price for TalTech

CPUcore

CPUcore*hour

0.008 EUR

GPU

GPU*hour

0.35 EUR

User home directory

20 GB

free

Project storage (persistent and scratch)

TB*hour

0.0106 EUR

Flash based scratch storage

TB*hour

10 x 0.0106 EUR

More detail guide how to calculate computational costs for LUMI can be found in LUMI billing policy.




SLURM partitions


partition

default time

time limit

default memory

nodes

short

10 min

4 hours

1 GB/thread

green, ada, amp

common

10 min

8 days

1 GB/thread

green

green-ib

10 min

8 days

1 GB/thread

green

long

10 min

22 days

1 GB/thread

green

gpu

10 min

5 days

1 GB/thread

amp, ada

bigmem

10 min

8 days

1 GB/thread

ada, amp, mem1tb




Contents: