اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدAn Empirical Analysis of VM Startup Times in Public IaaS Clouds: An Extended Report
55
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
VM startup time is an essential factor in designing elastic cloud applications. For example, a cloud application with autoscaling can reduce under- and over-provisioning of VM instances with a precise estimation of VM startup time, and in turn, it is likely to guarantee the applications performance and improve the cost efficiency. However, VM startup time has been little studied, and available measurement results performed previously did not consider various configurations of VMs for modern cloud applications. In this work, we perform comprehensive measurements and analysis of VM startup time from two major cloud providers, namely Amazon Web Services (AWS) and Google Cloud Platform (GCP). With three months of measurements, we collected more than 300,000 data points from each provider by applying a set of configurations, including 11+ VM types, four different data center locations, four VM image sizes, two OS types, and two purchase models (e.g., spot/preemptible VMs vs. on-demand VMs). With extensive analysis, we found that VM startup time can vary significantly because of several important factors, such as VM image sizes, data center locations, VM types, and OS types. Moreover, by comparing with previous measurement results, we confirm that cloud providers (specifically AWS) made significant improvements for the VM startup times and currently have much quicker VM startup times than in the past.
قيم البحث
اقرأ أيضاً
We have developed a highly scalable application, called Shoal, for tracking and utilizing a distributed set of HTTP web caches. Squid servers advertise their existence to the Shoal server via AMQP messaging by running Shoal Agent. The Shoal server pr
ovides a simple REST interface that allows clients to determine their closest Squid cache. Our goal is to dynamically instantiate Squid caches on IaaS clouds in response to client demand. Shoal provides the VMs on IaaS clouds with the location of the nearest dynamically instantiated Squid Cache. In this paper, we describe the design and performance of Shoal.
Internet of Things (IoT) has already proven to be the building block for next-generation Cyber-Physical Systems (CPSs). The considerable amount of data generated by the IoT devices needs latency-sensitive processing, which is not feasible by deployin
g the respective applications in remote Cloud datacentres. Edge/Fog computing, a promising extension of Cloud at the IoT-proximate network, can meet such requirements for smart CPSs. However, the structural and operational differences of Edge/Fog infrastructure resist employing Cloud-based service regulations directly to these environments. As a result, many research works have been recently conducted, focusing on efficient application and resource management in Edge/Fog computing environments. Scalable Edge/Fog infrastructure is a must to validate these policies, which is also challenging to accommodate in the real-world due to high cost and implementation time. Considering simulation as a key to this constraint, various software has been developed that can imitate the physical behaviour of Edge/Fog computing environments. Nevertheless, the existing simulators often fail to support advanced service management features because of their monolithic architecture, lack of actual dataset, and limited scope for a periodic update. To overcome these issues, we have developed multiple simulation models for service migration, dynamic distributed cluster formation, and microservice orchestration for Edge/Fog computing in this work and integrated with the existing iFogSim simulation toolkit for launching it as iFogSim2. The performance of iFogSim2 and its built-in policies are evaluated using three use case scenarios and compared with the contemporary simulators and benchmark policies under different settings. Results indicate that the proposed solution outperform others in service management time, network usage, ram consumption, and simulation time.
Data intensive applications on clusters often require requests quickly be sent to the node managing the desired data. In many applications, one must look through a sorted tree structure to determine the responsible node for accessing or storing the d
ata. Examples include object tracking in sensor networks, packet routing over the internet, request processing in publish-subscribe middleware, and query processing in database systems. When the tree structure is larger than the CPU cache, the standard implementation potentially incurs many cache misses for each lookup; one cache miss at each successive level of the tree. As the CPU-RAM gap grows, this performance degradation will only become worse in the future. We propose a solution that takes advantage of the growing speed of local area networks for clusters. We split the sorted tree structure among the nodes of the cluster. We assume that the structure will fit inside the aggregation of the CPU caches of the entire cluster. We then send a word over the network (as part of a larger packet containing other words) in order to examine the tree structure in another nodes CPU cache. We show that this is often faster than the standard solution, which locally incurs multiple cache misses while accessing each successive level of the tree.
We introduce an optimization algorithm for resource allocation in the LIPI Public Cluster to optimize its usage according to incoming requests from users. The tool is an extended and modified genetic algorithm developed to match specific natures of p
ublic cluster. We present a detail analysis of optimization, and compare the results with the exact calculation. We show that it would be very useful and could realize an automatic decision making system for public clusters.
Complex applications running on multicore processors show a rich performance phenomenology. The growing number of cores per ccNUMA domain complicates performance analysis of memory-bound code since system noise, load imbalance, or task-based programm
ing models can lead to thread desynchronization. Hence, the simplifying assumption that all cores execute the same loop can not be upheld. Motivated by observations on plain and modifi
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
