arXiv subjects

Pankaj Saha

Publications and source records attributed to Pankaj Saha.

26 records · Page 2Linked to original sources

arXiv2019

Tromino: Demand and DRF Aware Multi-Tenant Queue Manager for Apache Mesos Cluster

Apache Mesos, a two-level resource scheduler, provides resource sharing across multiple users in a multi-tenant cluster environment. Computational resources (i.e., CPU, memory, disk, etc. ) are distributed according to the Dominant Resource Fairness (DRF) policy. Mesos frameworks (users) receive resources based on their current usage and are responsible for scheduling their tasks within the allocation. We have observed that multiple frameworks can cause fairness imbalance in a multiuser environment. For example, a greedy framework consuming more than its fair share of resources can deny resource fairness to others. The user with the least Dominant Share is considered first by the DRF module to get its resource allocation. However, the default DRF implementation, in Apache Mesos' Master allocation module, does not consider the overall resource demands of the tasks in the queue for each user/framework. This lack of awareness can result in users without any pending task receiving more resource offers while users with a queue of pending tasks starve due to their high dominant shares. We have developed a policy-driven queue manager, Tromino, for an Apache Mesos cluster where tasks for individual frameworks can be scheduled based on each framework's overall resource demands and current resource consumption. Dominant Share and demand awareness of Tromino and scheduling based on these attributes can reduce (1) the impact of unfairness due to a framework specific configuration, and (2) unfair waiting time due to higher resource demand in a pending task queue. In the best case, Tromino can significantly reduce the average waiting time of a framework by using the proposed Demand-DRF aware policy.

cs.PF↗

arXiv2019

Evaluation of Docker Containers for Scientific Workloads in the Cloud

The HPC community is actively researching and evaluating tools to support execution of scientific applications in cloud-based environments. Among the various technologies, containers have recently gained importance as they have significantly better performance compared to full-scale virtualization, support for microservices and DevOps, and work seamlessly with workflow and orchestration tools. Docker is currently the leader in containerization technology because it offers low overhead, flexibility, portability of applications, and reproducibility. Singularity is another container solution that is of interest as it is designed specifically for scientific applications. It is important to conduct performance and feature analysis of the container technologies to understand their applicability for each application and target execution environment. This paper presents a (1) performance evaluation of Docker and Singularity on bare metal nodes in the Chameleon cloud (2) mechanism by which Docker containers can be mapped with InfiniBand hardware with RDMA communication and (3) analysis of mapping elements of parallel workloads to the containers for optimal resource management with container-ready orchestration tools. Our experiments are targeted toward application developers so that they can make informed decisions on choosing the container technologies and approaches that are suitable for their HPC workloads on cloud infrastructure. Our performance analysis shows that scientific workloads for both Docker and Singularity based containers can achieve near-native performance. Singularity is designed specifically for HPC workloads. However, Docker still has advantages over Singularity for use in clouds as it provides overlay networking and an intuitive way to run MPI applications with one container per rank for fine-grained resources allocation.

cs.PF↗

arXiv2019

Scylla: A Mesos Framework for Container Based MPI Jobs

Open source cloud technologies provide a wide range of support for creating customized compute node clusters to schedule tasks and managing resources. In cloud infrastructures such as Jetstream and Chameleon, which are used for scientific research, users receive complete control of the Virtual Machines (VM) that are allocated to them. Importantly, users get root access to the VMs. This provides an opportunity for HPC users to experiment with new resource management technologies such as Apache Mesos that have proven scalability, flexibility, and fault tolerance. To ease the development and deployment of HPC tools on the cloud, the containerization technology has matured and is gaining interest in the scientific community. In particular, several well known scientific code bases now have publicly available Docker containers. While Mesos provides support for Docker containers to execute individually, it does not provide support for container inter-communication or orchestration of the containers for a parallel or distributed application. In this paper, we present the design, implementation, and performance analysis of a Mesos framework, Scylla, which integrates Mesos with Docker Swarm to enable orchestration of MPI jobs on a cluster of VMs acquired from the Chameleon cloud [1]. Scylla uses Docker Swarm for communication between containerized tasks (MPI processes) and Apache Mesos for resource pooling and allocation. Scylla allows a policy-driven approach to determine how the containers should be distributed across the nodes depending on the CPU, memory, and network throughput requirement for each application.

cs.PF↗

arXiv2019

Minimal inflationary cosmologies and constraints on reheating

With the growing consensus on simple power law inflation models not being favored by the PLANCK observation, dynamics for the non-standard form of the inflaton potential gain significant interest in the recent past. In this paper, we analyze in great detail classes of phenomenologically motivated inflationary models with non-polynomial potential which are the generalization of the potential introduced in \cite{mhiggs}. After the end of inflation, inflaton field will coherently oscillate around its minimum. Depending upon the initial amplitude of the oscillation and coupling parameters standard parametric resonance phenomena will occur. Therefore, we will study how the inflationary model parameters play an important role in understanding the resonant structure of our model under study. Subsequently, the universe will go through the perturbative reheating phase. However, without any specific model consideration, we further study the constraints on our models based on model independent reheating constraint analysis.

astro-ph.CO↗

arXiv2019

Minimal plateau inflationary cosmologies and constraints from reheating

With the growing consensus on simple power law inflation models not being favored by the PLANCK observations, dynamics for the non-standard inflation gain significant interest in the recent past. In this paper, we analyze in detail a class of supergravity inspired phenomenological inflationary models with non-polynomial potential based on Maity (Nucl.Phys. B919 (2017) 560), and compare the model predictions with the currently most favored Starobinsky and its generalized $α$-attractor models in the ($n_s$ , $r$) plane constrained by PLANCK. Importantly for a wide range of parameter space, our model provides successful inflation in the sub-Planckian regime. We also have performed model independent analysis of reheating in terms of the effective equation of state parameter. In particular, we consider two stages of reheating dynamics with generalized inflaton equation of state in the initial and relativistic equation of state in the later phase. Finally, we show how our generalized reheating analysis constrains the inflation models under consideration

gr-qc↗

arXiv2018

Connecting CMB anisotropy and cold dark matter phenomenology via reheating

Understanding the properties of dark matter has proved to be one of the most challenging problems of particle phenomenology. In this paper, we have tried to understand the phenomenology of dark matter in light of very well understood properties of cosmic microwave background (CMB) anisotropy. To connect these two, inflation and its subsequent evolution known as the reheating phase play the important role. Following the previous analysis, we first established a one-to-one correspondence between the CMB power spectrum and the reheating temperature assuming the perturbative reheating scenario. Further by incorporating a possible dark matter candidate through the radiation annihilation process during reheating and the current value of dark matter abundance, we constrain the dark matter parameter space through the inflationary power spectrum for different inflationary models.

hep-ph↗

arXiv2018

Studying G-axion Inflation model in light of PLANCK

With the Planck 2015 result, most of the well known canonical large field inflation models turned out to be strongly disfavored. Axion inflation is one of such models which is becoming marginalized with the increasing precession of CMB data. In this paper, we have shown that with a simple Galileon type modification to the marginally favored axion model calling G-axion, we can turn them into one of the most favored models with its detectable prediction of $r$ and $n_s$ within its PLANCK $1σ$ range for a wide range of parameters. Interestingly it is this modification which plays the important role in turning the inflationary predictions to be independent of the explicit value of axion decay constant $f$. However, dynamics after the inflation turned out to have a non-trivial dependence on $f$. For each G-axion model there exists a critical value of $f_c$ such that for $f>f_c$ we have the oscillating phase after inflation and for $f<f_c$ we have non-oscillatory phase. Therefore, we obtained a range of sub-Planckian value of model parameters which give rise to consistent inflation. However for sub-Planckian axion decay constant the inflaton field configuration appeared to be singular after the end of inflation. To reheat the universe we, therefore, employ the instant preheating mechanism at the instant of first zero crossing of the inflaton. To our surprise, the instant preheating mechanism turned out to be inefficient as opposed to usual non-oscillatory quintessence model. For another class of G-axion model with super-Planckian axion decay constant, we performed in detail the reheating constraints analysis considering the latest PLANCK result.

hep-ph↗

arXiv2015

Modified natural inflation: A small single field model with a large tensor to scalar ratio

In this paper we explored in detail a phenomenological model of modified single field natural inflation in light of recent cosmological experiments, BICEP2. Our main goal is to construct an inflationary model which not only predicts the important cosmological quantities such as $(n_s, r)$ compatible with experimental observation, but also is consistent with the low energy effective theory framework. Therefore, all the fundamental scale apart from $M_p$ and quantities of our interest should be within the sub-Planckian region. In order to achieve our goal we modify the usual single field natural inflationary model by a specific form of higher derivative kinetic term called kinetic gravity braiding (KGB). One of our guiding principles to construct such a model is the constant shift symmetry of the axion. We have chosen the form of the KGB term in such a way that it predicts the required value of $n_s\simeq 0.96$ and a large tensor to scalar ratio $r> 0.1$. Importantly for a wide range of parameter space our model has sub-Planckian axion decay constant $f$ and the scale of inflation $Λ$. However, the reheating after the end of inflation limits the value of $f$ so that we are unable to get $f$ to be significantly lower than $M_p$. Furthermore, we find sub-Planckian field excursion for the axion field $Δϕ\simeq f$ for the sufficient number of e-folding ${\cal N} \gtrsim 50 $. We also discussed in detail about the natural preheating mechanism in our model based on the recently proposed Chern-Simon coupling. We found this gravity mediated preheating is very difficult to achieve in our model. With our general analytic argument, we also would like to emphasize that Chern-Simons mediated preheating is very unlikely to happen in any slow roll inflationary model.

hep-th↗