Memory sharing for handling memory overload on physical machines in cloud data centers

Abstract

Over-committing computing resources is a widely adopted strategy for increased cluster utilization in Infrastructure as a Service (IaaS) cloud data centers. A potential consequence of over-committing computing resources is memory overload of physical machines (PMs). Memory overload occurs if memory usage exceeds a defined alarm threshold, exposing running computation tasks at a risk of being terminated by the operating system. A prevailing measure to handle memory overload of a PM is live migration of virtual machines (VMs). However, this not only consumes network bandwidth, CPU, and other resources, but also compels a temporary unavailability of the VMs being migrated. To handle memory overload, we present a memory sharing system in this paper for PMs in cloud data centers. With memory sharing, a PM automatically borrows memory from a remote PM when necessary, and releases the borrowed memory when memory overload disappears. This is implemented through swapping inactive memory pages to remote memory resource. Experimental studies conducted on InfiniBand-networked PMs show that the memory sharing system is fully functional. The measured throughput and latency are around 929 Mbps and 1.3 $$\mu$$ μ s, respectively, on average for remote memory access. They are similar to those from accessing a local-volatile memory express solid-state drive, and thus are promising in real applications.

Construction of an Event Knowledge Graph Based on a Dynamic Resource Scheduling Optimization Algorithm and Semantic Graph Convolutional Neural Networks

“…Virtualization, as a method of provisioning resources, abstracts and transforms various physical resources (Servers, Storage, Networks, etc.). To run numerous virtual machines on physical servers, we employ KVM-QEMU virtualization technology and Kubernetes container technology [16]. These servers then help with big computer jobs like graph computation and knowledge triple extraction [17].…”

Joint Optimization of Memory Sharing and Communication Distance for Virtual Machine Instantiation in Cloudlet Networks

“…Ge et al [14] proposed a memory sharing system, which integrates a mechanism of threshold-based memory overload detection, that is presented for handling memory overload of InfiniBand-networked PMs in data centers. It enables a PM with memory overload to automatically borrow memory from a remote PM with spare memory.…”

“…p qv is the probability that q appears in the coverage area of v during the period ∆, and d ov is the communication distance between o and v. The objective function f 2 must satisfy partial constraints together with the objective function f 1 , that is, (9)-( 13). Equation (14) requires that when a UD appears in any BS's coverage range, this UD's VM should be instantiated on a cloudlet. Equation (15) requires that y qvo can only be set to 1 when the q's VM is instantiated in cloudlet o, i.e., x qo = 1; otherwise, y qvo can only be set to 0 when x qo = 0.…”

“…We define a local variable maxZ and use i to represent the iteration order instead of θ in (1), and we set it to 1. Then, in each iteration, we repeatedly calculate Z i jk for each pair of VMs and cloudlets, and we determine the cloudlet o k and the VM t j with the maximum Z i jk (lines [14][15][16][17][18][19]. When calculating the ratio of required resources for instantiating t j to the remaining resources of o k in line 17, we consider the impact of a i jk , as there are already VMs instantiated in o k .…”