NASA's HPC infrastructure is in need of serious modernization, since in its current form it is not able to meet the needs of organizations within the US National Aeronautics and Space Administration. As reported by The Register, this was the conclusion reached during the audit by the Office of the Inspector General.
It is noted that NASA's LPC infrastructure is obsolete and cannot effectively support modern workloads. For example, NASA's Advanced Supercomputing Center uses 18 thousand CPUs and only 48 GPU-based accelerators.
In addition, current computing power is not enough for all consumers. Therefore, some NASA departments and research centers are forced to purchase their own equipment and create a local LDC infrastructure. In particular, the Space Launch System team alone spends $250 thousand annually for these purposes instead of connecting to a centralized system. Virtually every division of NASA, with the exception of the Goddard Space Flight Center and the Stennis Space Center, has its own independent computing infrastructure.
Another reason for deploying on-premises capacity is the confusion around NASA cloud resources and policies, which makes it difficult to plan and estimate financial costs. The audit also showed that there are questions about the safety of NASA's supercomputer park. For example, there is no proper monitoring of some systems to which foreign users have access.
In general, the observed picture leads to delays in the implementation of space missions and additional costs. To eliminate the shortcomings, NASA management was recommended to carry out a comprehensive reform of the LDC sector, including an inventory of assets, identification of technological gaps and cyber risks. It is also necessary to develop a strategy to improve the distribution of available computing power.
If you notice an error, select it with the mouse and press CTRL+ENTER. | Can you write better? We always welcome new authors.
Source: