ZIH-Info No. 183, October 2024
Table of contents
Editor: Jacqueline Papperitz
ZIH and ScaDS.AI at SC'24 in Atlanta
From November 17 to 22, ZIH and ScaDS.AI will present their current research in the fields of HPC, data analysis and AI at a joint booth at SC'24 in Atlanta / USA. The conference is the largest platform of the international HPC community, where scientific and technological innovations as well as new products are presented. SC'24 takes place under the motto "HPC creates" and will discuss key issues and new developments in HPC, AI, networking, storage technologies and analytics. The ZIH will present its resources and methodological work in HPC. In addition, demos of the new functions of the interactive performance analysis tool Vampir and tool developments such as MetricQ, lo2s and PIKA will be offered. ScaDS.AI Dresden/Leipzig will present a series of living lab demos, including the revised versions of the machine learning training software asanAI and JUmPER (JUpyter meets PERformance), which, in addition to the previous Score-P instrumentation, now allows the recording of system metrics for user code in Jupyter as well as the interactive visualization of these recordings. In addition, user challenges and an on-demand hackathon with asanAI will be offered at the stand. By participating in a European Passport Stamp Rally, the booth is taking part in a campaign by European HPC centers that focuses on their collaboration and capabilities. (Contact: Frank Winkler)
Expansion of the data center infrastructure
The Lehmann-Zentrum Rechenzentrum (LZR), which was inaugurated in May 2015, is currently being expanded in terms of the performance and security of supply of its infrastructure. Among others, this will address the increased requirements for IT services and high-performance computing of the ZIH, DLR and further housing customers. In the area of high-performance computing, the work includes, for example, increasing the capacity of the hot water cooling system and expanding the electrical power to 4.8 MW. In addition, the construction measures already presented in the last issue of ZIH-Info for feeding waste heat into the municipal district heating network are part of the expansion. Interruptions to the operation of high-performance computers lasting several days are unfortunately unavoidable during the expansion. However, together with the companies carrying out the work, we are doing our best to keep these periods as short as possible. Users will be informed separately about specific dates of service interruptions well in advance. (Contact: Dr. Daniel Hackenberg)
SMB service for research data
For larger amounts of research data up to 50 TB SMB shares for joint access to network resources can now be requested via the Self-Service Portal. For even larger amounts of data, please write a ticket to the Service Desk with a reference to the request. Parallel use of NFS and SMB on the same data is currently not possible for research data. (Contact: Dr. Michael Kluge)
Group management for Dyport and VPN
The group administrations of Dyport and VPN for the TUD data network will be switched from a separate administration to the central group administration in the Self-Service Portal at the end of September. This means that existing groups can also be used for central IP address allocation and VPN access, thus avoiding the need for parallel maintenance. (Contact: )
Young Talent Award for ZIH student Justin Bürger
The German Datacenter Association once again presented awards to young talents from the IT and data center sector as conclusion of this year's German Datacenter Conference on 3 September 2024. Among them was Justin Bürger, who wrote his research paper on the reproducibility of research data using a model repository at TU Dresden's ZIH. He succeeded in isolating the biological information content and parameterization from a computer model on which a study on embryonic development in vertebrates is based. Due to dependencies on software libraries at the time, the source code of the model of the study published in the journal PLoS Computational Biology in 2011 was no longer executable. Justin Bürger used the MorpheusML model description language developed in the ZIH Department Innovative Methods of Computing to store the data in a sustainable and machine-readable way. Various simulation programs can now reuse and edit the model. Justin Bürger has thus shown how research data in the field of computational biology can be separated from the source code of specific simulation software using a standardized model description language and made persistently accessible to other simulators via data centers with model repositories. He has also used Morpheus to qualitatively and quantitatively reproduce and confirm the originally published results. Further information: https: //tud.de/zih/news/justin-buerger-erhaelt-young-talent-award-gda. (Contact: Dr. Lutz Brusch)
Two-factor authentication completed
In November 2023, two-factor authentication (2FA) was put into operation at TU Dresden. The aim of this introduction was to ensure secure login, especially for online exams. Users who had not set up 2FA for their account within the required period were initially reminded at the beginning of this year and blocked in the next step if they failed to take action. At times, this led to a high volume of inquiries at the Service Desk. Since July, inquiries on this topic have once again become part of day-to-day business. Currently, everyone who actively uses an account is equipped with 2 FA. Specifically, this includes 95% of students, 87 % of employees (incl. UKD) and 66 % of guests. Accounts that are still missing are expected to be provided in the coming weeks as part of feedback or account extensions. The project is therefore considered to have been successfully completed and the project management was discharged in the CDIO Strategy Council. (Contact: ).
Softphone piloting: survey
Telephoning via softphone, i.e. via PC or smartphone, is currently being tested at the TUD as part of a pilot project. Besides technical issues, aspects of communication and documentation are also evaluated. After inviting all employees, 791 participants in the pilot phase were asked about their experience of the switch in July 2024. With a response rate of 35 %, the project team now has valid and informative results: According to these, around 70 % of respondents felt that the changeover was positive. The feedback from some of the responses in the comments field is particularly valuable. These important indications and findings will be used for the planned second survey, which will focus on the functions and quality of the service. (Contact: )
ZIH colloquium
On October 24, 2024, at 15:00, the monthly ZIH colloquium will take place in APB 1096 with Mr. Juan Jiménez-Sánchez from the Department of Mathematical Sciences of the Politecnico di Torino, Italy. The topic of the lecture is "A stochastic discrete mesoscopic simulator of tumor growth: biomarker discovery and therapy optimization". Further information: https: //tu-dresden.de/zih/kolloquium. (Contact: Dr. Hartmut Mix)
Special events
- Softphone consultation hour:
07.10., 14.10., and 21.10., 28.10.2024: 10:00–11:00 a.m. (online) - 07.10.– 08.10.: All-Hands-Meeting of the AI community in Germany
(ScaDS.AI Dresden/Leipzig) - Open Q&A session for users of the NHR@TUD computing clusters:
14.10. and 28.10.2024: 13:30–14:30 (online) -
21.10.2024, 14:50–16:20: Reboot the Future - Feminist Perspectives in Human Computer Interaction #3: Material Trajectories: Designing with Care
-
23.10.2024, 14:00–14:45: Which ZIH infrastructure can I use for secure storage and sustainable management of my research data? (online)
-
24.10., 09:00–13:00: NHR-Tutorial: Efficient use and analysis of deep learning frameworks on HPC resources
-
28.10.2024, 13:00–14:30: Reboot the Future - Feminist Perspectives in Human Computer Interaction #4: Between Subjectivity and Imposition: Power Dynamics in Data Annotation for Computer Vision
-
07.11.2024, 9:20–10:50 a.m: Onyx Basic Course (online)
-
07.11.2024, 10:00–15:00: ScaDS.AI Tutorial: Big Data Processing on HPC