Dynamic Memory Elasticity for On-Premises Clouds
Automated Oversubscription for Cloud Platforms
View on GitHub

CloudVAMP main features

Memory Usage Monitoring

Monitors the memory usage of the VMs to detect the amount of free memory to be stolen.

Dynamic Memory Allocation

Memory size is dynamically stolen/assigned to VMs via KVM. Automated vertical elasticity.

Oversubscription Enabler

Stolen memory is reported to the Cloud Management Platform to enable oversubscription.

Live Migration

Live migration across nodes is employed to prevent memory overload in the physical hosts.

Integrated with OpenNebula

Fully integrated with OpenNebula via extensible mechanisms. No source code modifications required.

Open Source

Developed in Python, hosted in GitHub and distributed under the Apache 2.0 License.

Dynamic Memory Resize for VMs

Cloud users tend to overestimate the memory requirements of their applications/services. Templates (e.g. 2 vCPUs and 2048 MB of RAM) typically represent upper bounds for application requirements, thus wasting memory allocated to the VM. Also, applications may show highly dynamic memory requirements. Wouldn't it be great that VMs where allocated the precise memory at every point of time to be used by their applications? CloudVAMP monitors actual memory usage and dynamically resizes the memory allocated to VMs, relying on the memory ballooning support provided by the KVM hypervisor. The free memory available from VMs is reclaimed (stolen) by CloudVAMP.

Enable Oversubscription

The stolen memory from the VMs is exposed as free memory to the Cloud Management Platform (CMP). The CMP's scheduler can decide to allocate additional VMs on the physical hosts. Therefore, CloudVAMP enables oversubscription for your on-premises Cloud. Oversubscription increases virtual server density, this being able to run more VMs per physical host. But what if applications (and in turn VMs) start increasing their memory usage? CloudVAMP dynamically increases the memory allocated to VMs. It also introduces preventive mechanisms to avoid memory overload in the physical hosts.

Prevent Memory Overload by Live Migration

Oversubscription may result in memory overload of the physical host if no preventive measures are considered. For that, CloudVAMP prevents memory overload via the live migration of VMs across the physical nodes, as supported both by KVM and OpenNebula. No downtime is introduced for VMs and memory overload is prevented, thus maintaining the Level of Service.

Learn More »

Tel: (+34) 963 87 70 07 Ext. 88254
Camino de Vera Road, Building 8B, Door N, 1st Floor,
Valencia City, Valencia 46022

Copyright © 2015, GRyCAP-I3M-UPV, Universitat Politècnica de València - 46022, Valencia, Spain