arXiv subjects

Guilherme Thompson

Publications and source records attributed to Guilherme Thompson.

4 recordsLinked to original sources

arXiv2021

A white-boxed ISSM approach to estimate uncertainty distributions of Walmart sales

We present our solution for the M5 Forecasting - Uncertainty competition. Our solution ranked 6\ts{th} out of 909 submissions across all hierarchical levels and ranked first for prediction at the finest level of granularity (product-store sales, i.e.\ SKUs). The model combines a multi-stage state-space model and Monte Carlo simulations to generate the forecasting scenarios (trajectories). Observed sales are modelled with negative binomial distributions to represent discrete over-dispersed sales. Seasonal factors are hand-crafted and modelled with linear coefficients that are calculated at the store-department level.

stat.AP↗

arXiv2017

Allocation Schemes of Ressources with Downgrading

We consider a server with large capacity delivering video files encoded in various resolutions. We assume that the system is under saturation in the sense that the total demand exceeds the server capacity $C$. In such a case, requests may be rejected. For the policies considered in this paper, instead of rejecting a video request, it is downgraded. When the occupancy of the server is above some value $C_0{<}C$, the server delivers the video at a minimal bit rate. The quantity $C_0$ is the bit rate adaptation threshold. For these policies, request blocking is thus replaced by bit rate adaptation. Under the assumptions of Poisson request arrivals and exponential service times, we show that, by rescaling the system, a process associated to the occupancy of the server converges to some limiting process whose invariant distribution is computed by a Wiener-Hopf technique. This allows us to derive an asymptotic expression of the key performance measure of such a policy, the equilibrium probability that a request is transmitted without downgrading. Numerical applications of these results are presented.

cs.NI↗

arXiv2016

Analysis of an offloading scheme for data centers in the framework of fog computing

In the context of fog computing, we consider a simple case when data centers are installed at the edge of the network and assume that if a request arrives at an overloaded data center, then it is forwarded to a neighboring data center with some probability. Data centers are assumed to have a large number of servers and that traffic at some of them is causing saturation. In this case the other data centers may help to cope with this saturation regime by accepting some of the rejected requests. Our aim is to qualitatively estimate the gain achieved via cooperation between neighboring data centers. After proving some convergence results, related to the scaling limits of loss systems, for the process describing the number of free servers at both data centers, we show that the performance of the system can be expressed in terms of the invariant distribution of a random walk in the quarter plane. By using and developing existing results in the technical literature, explicit formulas for the blocking rates of such a system are derived.

cs.NI↗

arXiv2016

Analysis of a trunk reservation policy in the framework of fog computing

We analyze in this paper a system composed of two data centers with limited capacity in terms of servers. When one request for a single server is blocked at the first data center, this request is forwarded to the second one. To protect the single server requests originally assigned to the second data center, a trunk reservation policy is introduced (i.e., a redirected request is accepted only if there is a sufficient number of free servers at the second data center). After rescaling the system by assuming that there are many servers in both data centers and high request arrival rates, we are led to analyze a random walk in the quarter plane, which has the particularity of having non constant reflecting conditions on one boundary of the quarter plane. Contrary to usual reflected random walks, to compute the stationary distribution of the presented random walk, we have to determine three unknown functions, one polynomial and two infinite generating functions. We show that the coefficients of the polynomial are solutions to a linear system. After solving this linear system, we are able to compute the two other unknown functions and the blocking probabilities at both data centers. Numerical experiments are eventually performed to estimate the gain achieved by the trunk reservation policy.

cs.NI↗