No abstract

Non-uniform memory access (NUMA) architectures pose numerous performance challenges for main-memory column-stores in scaling up analytics on modern multi-socket multi-core servers. A NUMA-aware execution engine needs a strategy for data placement and task scheduling that prefers fast local memory accesses over remote memory accesses, and avoids an imbalance of resource utilization, both CPU and memory bandwidth, across sockets. State-of-the-art systems typically use a static strategy that always partitions data across sockets, and always allows inter-socket task stealing. In this paper, we show that adapting data placement and task stealing to the workload can improve throughput by up to a factor of 4 compared to a static approach. We focus on highly concurrent workloads dominated by operators working on a single table or table group (copartitioned tables). Our adaptive data placement algorithm tracks the resource utilization of tasks, partitions of tables and table groups, and sockets. When a utilization imbalance across sockets is detected, the algorithm corrects it by moving or repartitioning tables. Also, inter-socket task stealing is dynamically disabled for memory-intensive tasks that could otherwise hurt performance.

To cope with individualization and the high costs of downtimes, modern production systems should be flexible, adaptable, and resilient. Multi-Agent Systems are suitable to address these requirements by decentralizing production systems. However, the agent paradigm is still not widely applied. One of the key reasons is that the agents’ knowledge bases had to be created manually, which is cumbersome, error-prone, and insufficiently standardized. Digital Twins have the potential to solve this issue, as they describe relevant information in a standardized way. This paper presents an approach to leveraging Digital Twins, i. e., the Asset Administration Shell, to realize Multi-Agent Systems in the production context. For this, a parser automatically extracts relevant information from the Digital Twins and initializes the individual agents in a Multi-Agent System, i. e., PADE.

In the Auslander-Reiten quiver of a representation-directed algebra several hammocks occur naturally; they begin at the projective cover of a simple module E and end in the corresponding injective hull. It is known that hammocks are Auslander-Reiten quivers of posets, so there is a poset corresponding to each simple module; it describes the set of modules having E as a composition factor. In this paper we show that this poset 5 decomposes canonically into a coideal 5 + and an ideal 5 " which can easily be described by vectorspace-categories corresponding to a one-point extension or a one-point coextension, respectively. In addition, we describe the simple modules for which 5 + and S~ are not comparable, and also those for which S + > S~.We also show how to use the results in order to prove for certain posets that they do not occur as posets corresponding to simple modules.