Performance Analysis of Network I/O Workloads in Virtualized Data Centers

Abstract

Abstract-Server consolidation and application consolidation through virtualization are key performance optimizations in cloud-based service delivery industry. In this paper, we argue that it is important for both cloud consumers and cloud providers to understand the various factors that may have significant impact on the performance of applications running in a virtualized cloud. This paper presents an extensive performance study of network I/O workloads in a virtualized cloud environment. We first show that current implementation of virtual machine monitor (VMM) does not provide sufficient performance isolation to guarantee the effectiveness of resource sharing across multiple virtual machine instances (VMs) running on a single physical host machine, especially when applications running on neighboring VMs are competing for computing and communication resources. Then we study a set of representative workloads in cloud-based data centers, which compete for either CPU or network I/O resources, and present the detailed analysis on different factors that can impact the throughput performance and resource sharing effectiveness. For example, we analyze the cost and the benefit of running idle VM instances on a physical host where some applications are hosted concurrently. We also present an in-depth discussion on the performance impact of colocating applications that compete for either CPU or network I/O resources. Finally, we analyze the impact of different CPU resource scheduling strategies and different workload rates on the performance of applications running on different VMs hosted by the same physical machine.

Cloud-enabled prognosis for manufacturing

“…Barham et al [18] pointed out that 30-40% of execution time for a network to transmit or receive an operation was spent in VMM to remap addresses contained in the transmitted data package. It has been demonstrated that the overhead of central processing units (CPUs) and latency increase with the transmitted package rate due to increased communication between the server (VMM) and client (VMs) domains [136].…”

PIASA: A power and interference aware resource management strategy for heterogeneous workloads in cloud data centers

“…Mars et al [24] propose that one way to measure the QoS of web servers is to measure the server throughput at different workload rates, namely, the maximum number of successful queries per second, as is the case in Google's web search. According to Mei et al [26], the performance of web servers is CPU bound under a mix of small files and is network bound under a mix of large files. This work considers the second case by assuming that web workloads are mainly characterized by network resource consumption (i.e., are network-intensive) and have residual CPU consumption.…”

“…This work considers the second case by assuming that web workloads are mainly characterized by network resource consumption (i.e., are network-intensive) and have residual CPU consumption. The CPU time consumed to process network requests can be divided into two major categories: the time spent establishing TCP connections and the time spent transporting web file content [26]. Because the demand for resources changes abruptly with time, the amount of transferred network I/O bðsÞ served by a task s at each instant is given by Eq.…”

A Model-Driven DevOps Framework for QoS-Aware Cloud Applications

“…The delay center is used to model network and/or protocol delays introduced in establishing connections, etc. Performance parameters need to be continuously updated at runtime (see [12] for further details) in order to capture transient behavior of VMs network and I/O interference [13], and performance variability of the Cloud provider over time [14]. This simplified performance model is used since alternative runtime adaptation decisions need to be evaluated very quickly according to a 5-10 minutes time scale.…”