-
Reinforcement Learning-based Application Autoscaling in the Cloud: A Survey
Authors:
Yisel Garí,
David A. Monge,
Elina Pacini,
Cristian Mateos,
Carlos García Garino
Abstract:
Reinforcement Learning (RL) has demonstrated a great potential for automatically solving decision-making problems in complex uncertain environments. RL proposes a computational approach that allows learning through interaction in an environment with stochastic behavior, where agents take actions to maximize some cumulative short-term and long-term rewards. Some of the most impressive results have…
▽ More
Reinforcement Learning (RL) has demonstrated a great potential for automatically solving decision-making problems in complex uncertain environments. RL proposes a computational approach that allows learning through interaction in an environment with stochastic behavior, where agents take actions to maximize some cumulative short-term and long-term rewards. Some of the most impressive results have been shown in Game Theory where agents exhibited superhuman performance in games like Go or Starcraft 2, which led to its gradual adoption in many other domains, including Cloud Computing. Therefore, RL appears as a promising approach for Autoscaling in Cloud since it is possible to learn transparent (with no human intervention), dynamic (no static plans), and adaptable (constantly updated) resource management policies to execute applications. These are three important distinctive aspects to consider in comparison with other widely used autoscaling policies that are defined in an ad-hoc way or statically computed as in solutions based on meta-heuristics. Autoscaling exploits the Cloud elasticity to optimize the execution of applications according to given optimization criteria, which demands to decide when and how to scale-up/down computational resources, and how to assign them to the upcoming processing workload. Such actions have to be taken considering that the Cloud is a dynamic and uncertain environment. Motivated by this, many works apply RL to the autoscaling problem in the Cloud. In this work, we survey exhaustively those proposals from major venues, and uniformly compare them based on a set of proposed taxonomies. We also discuss open problems and prospective research in the area.
△ Less
Submitted 17 November, 2020; v1 submitted 27 January, 2020;
originally announced January 2020.
-
CMI: An Online Multi-objective Genetic Autoscaler for Scientific and Engineering Workflows in Cloud Infrastructures with Unreliable Virtual Machines
Authors:
David A. Monge,
Elina Pacini,
Cristian Mateos,
Enrique Alba,
Carlos García Garino
Abstract:
Cloud Computing is becoming the leading paradigm for executing scientific and engineering workflows. The large-scale nature of the experiments they model and their variable workloads make clouds the ideal execution environment due to prompt and elastic access to huge amounts of computing resources. Autoscalers are middleware-level software components that allow scaling up and down the computing pl…
▽ More
Cloud Computing is becoming the leading paradigm for executing scientific and engineering workflows. The large-scale nature of the experiments they model and their variable workloads make clouds the ideal execution environment due to prompt and elastic access to huge amounts of computing resources. Autoscalers are middleware-level software components that allow scaling up and down the computing platform by acquiring or terminating virtual machines (VM) at the time that workflow's tasks are being scheduled. In this work we propose a novel online multi-objective autoscaler for workflows denominated Cloud Multi-objective Intelligence (CMI), that aims at the minimization of makespan, monetary cost and the potential impact of errors derived from unreliable VMs. In addition, this problem is subject to monetary budget constraints. CMI is responsible for periodically solving the autoscaling problems encountered along the execution of a workflow. Simulation experiments on four well-known workflows exhibit that CMI significantly outperforms a state-of-the-art autoscaler of similar characteristics called Spot Instances Aware Autoscaling (SIAA). These results convey a solid base for deepening in the study of other meta-heuristic methods for autoscaling workflow applications using cheap but unreliable infrastructures.
△ Less
Submitted 2 November, 2018;
originally announced November 2018.
-
Evaluating the importance of environmental factors on the spatial distribution of livestock settlements in the Monte desert with a Monte Carlo based model: Settlement Dynamics in Drylands (SeDD)
Authors:
Emmanuel N. Millán,
Silvana Goirán,
Julieta N. Aranibar,
Leonardo Forconesi,
Carlos García Garino,
Eduardo M. Bringa
Abstract:
In the Monte desert, increasing population density, changing land rights and infrastructure may encourage livestock activity, with unknown consequences on ecosystems. Factors that influence livestock settlement distribution may affect ecosystem degradation. We hypothesize that surface and groundwater availability influence livestock settlements distribution. We evaluated this hypothesis with a Mon…
▽ More
In the Monte desert, increasing population density, changing land rights and infrastructure may encourage livestock activity, with unknown consequences on ecosystems. Factors that influence livestock settlement distribution may affect ecosystem degradation. We hypothesize that surface and groundwater availability influence livestock settlements distribution. We evaluated this hypothesis with a Monte Carlo based model to simulate Settlement Dynamics in Drylands (SeDD), which calculates probabilities on a gridded region based on six environmental factors: groundwater depth, vegetation type, proximity to rivers, paved roads, old river beds, and existing settlements. A parameter sweep, including millions of simulations, was run with combinations of parameters related to these factors. The sets of parameter values that minimized the residuals between simulations and observations indicated the relative importance of each factors on settlement distribution. Distances to rivers and old river beds were critical to explain the current distribution of settlements, while vegetation, paved roads, and water table depths were not important. Spatial distribution of simulated vegetation, which included degradation around livestock settlements, generally agreed with remotely sensed vegetation classes. The model could be a useful tool to evaluate the effects of land use changes (water provision, river flows), on settlement distribution and vegetation degradation in arid environments.
△ Less
Submitted 28 July, 2015;
originally announced July 2015.
-
In-Medium rho^0 Spectral Function Study via the 2H, 3He, 12C(gamma,pi+pi-) Reaction
Authors:
G. M. Huber,
G. J. Lolos,
Z. Papandreou,
A. Shinozaki,
E. J. Brash,
M. Iurescu,
G. Garino,
K. Maruyama,
K. Maeda,
T. Suda,
A. Toyofuku,
B. K. Jennings,
A. Sasaki,
H. Yamashita
Abstract:
We report a helicity analysis of sub-threshold rho^0 production on 2H, 3He and 12C at low photo-production energies. The results are indicative of a large longitudinal rho^0 polarization (l=1, m=0) and are consistent with a strong helicity-flip production mechanism. This signature is used to extract in-medium rho^0_L invariant mass distributions for all three nuclei in a manner which is less mod…
▽ More
We report a helicity analysis of sub-threshold rho^0 production on 2H, 3He and 12C at low photo-production energies. The results are indicative of a large longitudinal rho^0 polarization (l=1, m=0) and are consistent with a strong helicity-flip production mechanism. This signature is used to extract in-medium rho^0_L invariant mass distributions for all three nuclei in a manner which is less model-dependent than previous measurements. The results are compared to kinematic and phenomenological models of the rho^0 spectral function. The 2H and 3He data distributions support the role of N*(1520) excitation in sha** the in-medium rho^0_L invariant mass distribution, while the 12C distributions are consistent with quasi-free rho^0_L production. The data support an in-medium modification of the rho^0_L invariant mass distribution.
△ Less
Submitted 10 October, 2003;
originally announced October 2003.
-
Helicity Signatures in Subthreshold rho Production on Nuclei
Authors:
G. J. Lolos,
G. M. Huber,
A. Shinozaki,
Z. Papandreou,
E. J. Brash,
K. Hossain,
M. Iurescu,
D. Nordin,
G. Garino,
K. Maruyama,
K. Maeda,
T. Suda,
A. Toyofuku,
A. Sasaki,
H. Yamashita
Abstract:
We report a helicity analysis of subthreshold rho production on 2H and 12C nuclei at low photoproduction energies and large |t|. The results are indicative of a large longitudinal rho polarization (l=1, m=0) and are consistent with a strong helicity-flip mechanism of rho production. The analysis is model-independent and supports the large mass shift reported from the earlier 3He experiments.
We report a helicity analysis of subthreshold rho production on 2H and 12C nuclei at low photoproduction energies and large |t|. The results are indicative of a large longitudinal rho polarization (l=1, m=0) and are consistent with a strong helicity-flip mechanism of rho production. The analysis is model-independent and supports the large mass shift reported from the earlier 3He experiments.
△ Less
Submitted 29 March, 2001;
originally announced March 2001.
-
Probing the DeltaNN component of 3He
Authors:
G. M. Huber,
G. J. Lolos,
E. J. Brash,
S. Dumalski,
F. Farzanpay,
M. Iurescu,
Z. Papandreou,
A. Shinozaki,
A. Weinerman,
T. Emura,
H. Hirosawa,
K. Niwa,
H. Yamashita,
K. Maeda,
T. Terasawa,
H. Yamazaki,
S. Endo,
K. Miyamoto,
Y. Sumi,
G. Garino,
K. Maruyama,
A. Leone,
R. Perrino,
T. Maki,
A. Sasaki
, et al. (1 additional authors not shown)
Abstract:
The 3He(gamma,pi^+/- p) reactions were measured simultaneously over a tagged photon energy range of 800<E_gamma<1120 MeV, well above the Delta resonance region. An analysis was performed to kinematically isolate Delta knockout events from conventional Delta photoproduction events, and a statistically significant excess of pi+p events was identified, consistent with Delta++ knockout. Two methods…
▽ More
The 3He(gamma,pi^+/- p) reactions were measured simultaneously over a tagged photon energy range of 800<E_gamma<1120 MeV, well above the Delta resonance region. An analysis was performed to kinematically isolate Delta knockout events from conventional Delta photoproduction events, and a statistically significant excess of pi+p events was identified, consistent with Delta++ knockout. Two methods were used to estimate the DeltaNN probability in the 3He ground state, corresponding to the observed knockout cross section. The first gave a lower probability limit of 1.5+/-0.6+/-0.5%; the second yielded an upper limit of about 2.6%.
△ Less
Submitted 14 April, 2000; v1 submitted 1 December, 1999;
originally announced December 1999.
-
Subthreshold rho^0 photoproduction on 3He
Authors:
TAGX Collaboration,
M. A. Kagarlis,
Z. Papandreou,
G. M. Huber,
G. J. Lolos,
A. Shinozaki,
E. J. Brash,
F. Farzanpay,
M. Iurescu,
A. Weinerman,
G. Garino,
K. Maruyama,
O. Konno,
K. Maeda,
T. Terasawa,
H. Yamazaki,
T. Emura,
H. Hirosawa,
K. Niwa,
H. Yamashita,
S. Endo,
K. Miyamoto,
Y. Sumi,
A. Leone,
R. Perrino
, et al. (3 additional authors not shown)
Abstract:
A large reduction of the rho^0 mass in the nuclear medium is reported, inferred from dipion photoproduction spectra in the 1 GeV region, for the reaction 3He(gamma,pi+ pi-)X with a 10% duty factor tagged-photon beam and the TAGX multi-particle spectrometer. The energy range covered (800 < E(gamma) < 1120 MeV) lies mostly below the free rho^0 production threshold, a region which is believed sensi…
▽ More
A large reduction of the rho^0 mass in the nuclear medium is reported, inferred from dipion photoproduction spectra in the 1 GeV region, for the reaction 3He(gamma,pi+ pi-)X with a 10% duty factor tagged-photon beam and the TAGX multi-particle spectrometer. The energy range covered (800 < E(gamma) < 1120 MeV) lies mostly below the free rho^0 production threshold, a region which is believed sensitive to modifications of light vector-meson properties at nuclear-matter densities. The rho^0 masses extracted from the MC fitting of the data, m*(rho^0) = 642 +/- 40, 669 +/- 32, and 682 +/- 56 MeV/c^2 for E(gamma) in the 800-880, 880-960, and 960-1040 MeV regions respectively, are independently corroborated by a measured, assumption-free, kinematical observable. This mass shift, far exceeding current mean-field driven theoretical predictions, may be suggestive of rho^0 decay within the range of the nucleonic field.
△ Less
Submitted 23 November, 1998;
originally announced November 1998.
-
Measurement of the Electric and Magnetic Polarizabilities of the Proton
Authors:
B. E. MacGibbon,
G. Garino,
M. A. Lucas,
A. M. Nathan,
G. Feldman,
B. Dolbilkin
Abstract:
The Compton scattering cross section on the proton has been measured at laboratory angles of 90$^\circ$ and 135$^\circ$ using tagged photons in the energy range 70--100 MeV and simultaneously using untagged photons in the range 100--148~MeV. With the aid of dispersion relations, these cross sections were used to extract the electric and magnetic polarizabilities, $\barα$ and $\barβ$ respectively…
▽ More
The Compton scattering cross section on the proton has been measured at laboratory angles of 90$^\circ$ and 135$^\circ$ using tagged photons in the energy range 70--100 MeV and simultaneously using untagged photons in the range 100--148~MeV. With the aid of dispersion relations, these cross sections were used to extract the electric and magnetic polarizabilities, $\barα$ and $\barβ$ respectively, of the proton. We find
$$\barα+\barβ =
( 15.0 \pm 2.9 \pm 1.1 \pm 0.4 ) \times 10^{-4} \: {\rm fm}^3,$$
in agreement with a model-independent dispersion sum rule, and
$$\barα-\barβ =
( 10.8 \pm 1.1 \pm 1.4 \pm 1.0 ) \times 10^{-4} \: {\rm fm}^3,$$
where the errors shown are statistical, systematic, and model-dependent, respectively. A comparison with previous experiments is given and global values for the polarizabilities are extracted.
△ Less
Submitted 10 July, 1995;
originally announced July 1995.
