Abstract
| The major upgrade of the ALICE experiment for the LHC Run3 poses unique challenges and opportunities for new software development. In particular, the entirely new data taking and processing software of ALICE relies on process parallelism and large amounts of shared objects in memory. Thus from a single-core single thread workload in the past, the new workloads are exclusively multithreaded. This requires a profound change in the ALICE Grid midleware job handling, from scheduling to execution, and thus the entire middleware has been rewritten during the past 3 years to support the new multithreaded reality.This paper presents the ALICE middlewre development for multi-core job management and the tools used to achieve an efficient and secure environment. In particular, it covers job isolation and scheduling and how they can be implemented in different site configurations, such as sites shared with other experiments or High Performance Computing resources. |