Garbage collection optimization for non uniform memory access architectures

Cache-coherent non uniform memory access (ccNUMA) architecture is a standard design pattern for contemporary multicore processors, and future generations of architectures are likely to be NUMA. NUMA architectures create new challenges for managed runtime systems. Memory-intensive applications use the system’s distributed memory banks to allocate data, and the automatic memory manager collects garbage left in these memory banks. The garbage collector may need to access remote memory banks, which entails access latency overhead and potential bandwidth saturation for the interconnection between memory banks. This dissertation makes five significant contributions to garbage collection on NUMA systems, with a case study implementation using the Hotspot Java Virtual Machine. It empirically studies data locality for a Stop-The-World garbage collector when tracing connected objects in NUMA heaps. First, it identifies a locality richness which exists naturally in connected objects that contain a root object and its reachable set— ‘rooted sub-graphs’. Second, this dissertation leverages the locality characteristic of rooted sub-graphs to develop a new NUMA-aware garbage collection mechanism. A garbage collector thread processes a local root and its reachable set, which is likely to have a large number of objects in the same NUMA node. Third, a garbage collector thread steals references from sibling threads that run on the same NUMA node to improve data locality. This research evaluates the new NUMA-aware garbage collector using seven benchmarks of an established real-world DaCapo benchmark suite. In addition, evaluation involves a widely used SPECjbb benchmark and Neo4J graph database Java benchmark, as well as an artificial benchmark. The results of the NUMA-aware garbage collector on a multi-hop NUMA architecture show an average of 15% performance improvement. Furthermore, this performance gain is shown to be as a result of an improved NUMA memory access in a ccNUMA system. Fourth, the existing Hotspot JVM adaptive policy for configuring the number of garbage collection threads is shown to be suboptimal for current NUMA machines. The policy uses outdated assumptions and it generates a constant thread count. In fact, the Hotspot JVM still uses this policy in the production version. This research shows that the optimal number of garbage collection threads is application-specific and configuring the optimal number of garbage collection threads yields better collection throughput than the default policy. Fifth, this dissertation designs and implements a runtime technique, which involves heuristics from dynamic collection behavior to calculate an optimal number of garbage collector threads for each collection cycle. The results show an average of 21% improvements to the garbage collection performance for DaCapo benchmarks.

Managing university records and documents in the world of governance, audit and risk: Case studies from South Africa and Malawi

There are enormous benefits for any organisation from practising sound records management. In the context of a public university, the importance of good records management includes: facilitating the achievement the university’s mandate; enhancing efficiency of the university; maintaining a reliable institutional memory; promoting trust; responding to an audit culture; enhancing university competitiveness; supporting the university’s fiduciary duty; demonstrating transparency and accountability; and fighting corruption. Records scholars and commentators posit that effective recordkeeping is an essential underpinning of good governance. Although there is a portrayal of positive correlation, recordkeeping struggles to get the same attention as that given to the governance. Evidence abounds of cases of neglect of recordkeeping in universities and other institutions in Sub-Saharan Africa. The apparent absence of sound recordkeeping provided a rationale for revisiting some universities in South Africa and Malawi in order to critically explore the place of recordkeeping in an organisation’s strategy in order to develop an alternative framework for managing records and documents in an era where good governance is a global agenda. The research is a collective case study in which multiple cases are used to critically explore the relationship between recordkeeping and governance. As qualitative research that belongs in the interpretive tradition of enquiry, it is not meant to suggest prescriptive solutions to general recordkeeping problems but rather to provide an understanding of the challenges and opportunities that arise in managing records and documents in the world of governance, audit and risk. That is: what goes on in the workplace; what are the problems; and what alternative approaches might address any existing problem situations. Research findings show that some institutions are making good use of their governance structures and other drivers for recordkeeping to put in place sound recordkeeping systems. Key governance structures and other drivers for recordkeeping identified include: laws and regulations; governing bodies; audit; risk; technology; reforms; and workplace culture. Other institutions are not managing their records and documents well despite efforts to improve their governance systems. They lack recordkeeping capacity. Areas that determine recordkeeping capacity include: availability of records management policy; capacity for digital records; availability of a records management unit; senior management support; level of education and training of records management staff; and systems and procedures for storage, retrieval and dispositions of records. Although this research reveals that the overall recordkeeping in the selected countries has slightly improved compared with the situation other researchers found a decade ago, it remains unsatisfactory and disjointed from governance. The study therefore proposes governance recordkeeping as an approach to managing records and documents in the world of governance, audit and risk. The governance recordkeeping viewpoint considers recordkeeping as a governance function that should be treated in the same manner as other governance functions such as audit and risk management. Additionally, recordkeeping and governance should be considered as symbiotic elements of a strategy. A strategy that neglects recordkeeping may not fulfil the organisation’s objectives effectively.

Physicochemical complexity in complex chemical systems

Self-replication and compartmentalization are two central properties thought to be essential for minimal life, and understanding how such processes interact in the emergence of complex reaction networks is crucial to exploring the development of complexity in chemistry and biology. Autocatalysis can emerge from multiple different mechanisms such as formation of an initiator, template self-replication and physical autocatalysis (where micelles formed from the reaction product solubilize the reactants, leading to higher local concentrations and therefore higher rates). Amphiphiles are also used in artificial life studies to create protocell models such as micelles, vesicles and oil-in-water droplets, and can increase reaction rates by encapsulation of reactants. So far, no template self-replicator exists which is capable of compartmentalization, or transferring this molecular scale phenomenon to micro or macro-scale assemblies. Here a system is demonstrated where an amphiphilic imine catalyses its own formation by joining a non-polar alkyl tail group with a polar carboxylic acid head group to form a template, which was shown to form reverse micelles by Dynamic Light Scattering (DLS). The kinetics of this system were investigated by 1H NMR spectroscopy, showing clearly that a template self-replication mechanism operates, though there was no evidence that the reverse micelles participated in physical autocatalysis. Active oil droplets, composed from a mixture of insoluble organic compounds in an aqueous sub-phase, can undergo processes such as division, self-propulsion and chemotaxis, and are studied as models for minimal cells, or protocells. Although in most cases the Marangoni effect is responsible for the forces on the droplet, the behaviour of the droplet depends heavily on the exact composition. Though theoretical models are able to calculate the forces on a droplet, to model a mixture of oils on an aqueous surface where compounds from the oil phase are dissolving and diffusing through the aqueous phase is beyond current computational capability. The behaviour of a droplet in an aqueous phase can only be discovered through experiment, though it is determined by the droplet's composition. By using an evolutionary algorithm and a liquid handling robot to conduct droplet experiments and decide which compositions to test next, entirely autonomously, the composition of the droplet becomes a chemical genome capable of evolution. The selection is carried out according to a fitness function, which ranks the formulation based on how well it conforms to the chosen fitness criteria (e.g. movement or division). Over successive generations, significant increases in fitness are achieved, and this increase is higher with more components (i.e. greater complexity). Other chemical processes such as chemiluminescence and gelation were investigated in active oil droplets, demonstrating the possibility of controlling chemical reactions by selective droplet fusion. Potential future applications for this might include combinatorial chemistry, or additional fitness goals for the genetic algorithm. Combining the self-replication and the droplet protocells research, it was demonstrated that the presence of the amphiphilic replicator lowers the interfacial tension between droplets of a reaction mixture in organic solution and the alkaline aqueous phase, causing them to divide. Periodic sampling by a liquid handling robot revealed that the extent of droplet fission increased as the reaction progressed, producing more individual protocells with increased self-replication. This demonstrates coupling of the molecular scale phenomenon of template self-replication to a macroscale physicochemical effect.