arXiv subjects

Julien Robert

Publications and source records attributed to Julien Robert.

24 records · Page 2Linked to original sources

arXiv2008

Algorithms for Temperature-Aware Task Scheduling in Microprocessor Systems

We study scheduling problems motivated by recently developed techniques for microprocessor thermal management at the operating systems level. The general scenario can be described as follows. The microprocessor's temperature is controlled by the hardware thermal management system that continuously monitors the chip temperature and automatically reduces the processor's speed as soon as the thermal threshold is exceeded. Some tasks are more CPU-intensive than other and thus generate more heat during execution. The cooling system operates non-stop, reducing (at an exponential rate) the deviation of the processor's temperature from the ambient temperature. As a result, the processor's temperature, and thus the performance as well, depends on the order of the task execution. Given a variety of possible underlying architectures, models for cooling and for hardware thermal management, as well as types of tasks, this scenario gives rise to a plethora of interesting and never studied scheduling problems. We focus on scheduling real-time jobs in a simplified model for cooling and thermal management. A collection of unit-length jobs is given, each job specified by its release time, deadline and heat contribution. If, at some time step, the temperature of the system is t and the processor executes a job with heat contribution h, then the temperature at the next step is (t+h)/2. The temperature cannot exceed the given thermal threshold T. The objective is to maximize the throughput, that is, the number of tasks that meet their deadlines. We prove that, in the offline case, computing the optimum schedule is NP-hard, even if all jobs are released at the same time. In the online case, we show a 2-competitive deterministic algorithm and a matching lower bound.

cs.DS↗

arXiv2006

Non-Clairvoyant Batch Sets Scheduling: Fairness is Fair enough

Scheduling questions arise naturally in many different areas among which operating system design, compiling,... In real life systems, the characteristics of the jobs (such as release time and processing time) are usually unknown and unpredictable beforehand. The system is typically unaware of the remaining work in each job or of the ability of the job to take advantage of more resources. Following these observations, we adopt the job model by Edmonds et al (2000, 2003) in which the jobs go through a sequence of different phases. Each phase consists of a certain quantity of work and a speed-up function that models how it takes advantage of the number of processors it receives. We consider the non-clairvoyant online setting where a collection of jobs arrives at time 0. We consider the metrics setflowtime introduced by Robert et al (2007). The goal is to minimize the sum of the completion time of the sets, where a set is completed when all of its jobs are done. If the input consists of a single set of jobs, this is simply the makespan of the jobs; and if the input consists of a collection of singleton sets, it is simply the flowtime of the jobs. We show that the non-clairvoyant strategy EQUIoEQUI that evenly splits the available processors among the still unserved sets and then evenly splits these processors among the still uncompleted jobs of each unserved set, achieves a competitive ratio (2+\sqrt3+o(1))\frac{ln n}{lnln n} for the setflowtime minimization and that this is asymptotically optimal (up to a constant factor), where n is the size of the largest set. For makespan minimization, we show that the non-clairvoyant strategy EQUI achieves a competitive ratio of (1+o(1))\frac{ln n}{lnln n}, which is again asymptotically optimal.

cs.DC↗

arXiv2006

Hierarchical geometric frustration in La3Cu2VO9

The crystallographic structure and magnetic properties of the La3Cu2VO9 were investigated by powder neutron diffraction and magnetization measurements. The compound materializes geometric frustration at two spatial scales, within clusters and between clusters, and at different temperature scales. It is shown by exactly solving the hamiltonian spectrum that collective spins are formed on each clusters at low temperature before inter-clusters coupling operates.

cond-mat.str-el↗

arXiv2006

Pull-Based Data Broadcast with Dependencies: Be Fair to Users, not to Items

Broadcasting is known to be an efficient means of disseminating data in wireless communication environments (such as Satellite, mobile phone networks,...). It has been recently observed that the average service time of broadcast systems can be considerably improved by taking into consideration existing correlations between requests. We study a pull-based data broadcast system where users request possibly overlapping sets of items; a request is served when all its requested items are downloaded. We aim at minimizing the average user perceived latency, i.e. the average flow time of the requests. We first show that any algorithm that ignores the dependencies can yield arbitrary bad performances with respect to the optimum even if it is given arbitrary extra resources. We then design a $(4+ε)$-speed $O(1+1/ε^2)$-competitive algorithm for this setting that consists in 1) splitting evenly the bandwidth among each requested set and in 2) broadcasting arbitrarily the items still missing in each set into the bandwidth the set has received. Our algorithm presents several interesting features: it is simple to implement, non-clairvoyant, fair to users so that no user may starve for a long period of time, and guarantees good performances in presence of correlations between user requests (without any change in the broadcast protocol). We also present a $ (4+ε)$-speed $O(1+1/ε^3)$-competitive algorithm which broadcasts at most one item at any given time and preempts each item broadcast at most once on average. As a side result of our analysis, we design a competitive algorithm for a particular setting of non-clairvoyant job scheduling with dependencies, which might be of independent interest.

cs.DS↗

arXiv2006

Spin liquid correlations in Nd-langasite anisotropic Kagomé antiferromagnet

Dynamical magnetic correlations in the geometrically frustrated Nd$\_3$Ga$\_5$SiO$\_{14}$ compound were probed by inelastic neutron scattering on a single crystal. A scattering signal with a ring shape distribution in reciprocal space and unprecedented dispersive features was discovered. Comparison with calculated static magnetic scattering from models of correlated spins suggests that the observed phase is a spin liquid inherent to an antiferromagnetic kagomé-like lattice of anisotropic Nd moments.

cond-mat.str-el↗

arXiv2006

Magnetic excitations in a new anisotropic Kagomé antiferromagnet

The Nd-langasite compound contains planes of magnetic Nd3+ ions on a lattice topologically equivalent to a kagomé net. The magnetic susceptibility does not reveal any signature of long-range ordering down to 2 K but rather a correlated paramagnetism with significant antiferromagnetic interactions between the Nd and a single-ion anisotropy due to crystal field effect. Inelastic neutron scattering on Nd-langasite powder and single-crystal allowed to probe its very peculiar low temperature dynamical magnetic correlations. They present unusual dispersive features and are broadly localized in wave-vector Q revealing a structure factor associated to characteristics short range-correlations between the magnetic atoms. From comparison with theoretical calculations, these results are interpreted as a possible experimental observation of a spin liquid state in an anisotropic kagomé antiferromagnet.

cond-mat.str-el↗