Information Technology Reference
In-Depth Information
specii cally address the four technical challenges as discussed previously.
A DVM has three key functionalities: (1) VE modeling, (2) service provision-
ing, and (3) mobility. In DVM, a VM and its model are integrated to form a
(integrated) VE. A VM model dei ned in an integrated VE is named a DVM.
The DVM is a primary building block of an integrated VE and a fundamental
unit of management. A VE can be a standalone VM or a networked distrib-
uted virtual environment with several VMs connected by overlay virtual
networks. The integrated VE contains all virtual hardware and software
idiosyncrasies and their coni gurations. A VE can be coni gured and recon-
i gured based on its model, which provides the means of VE coni guration
management. Within a VE, a VM or a set of VMs can be remodeled and rede-
ployed, which provides the means of system coni guration management.
Through VM modeling and provisioning, a DVM provides a thorough solu-
tion to system coni guration management. The VM modeling and service
provisioning functions support basic VE lifetime managements. VM model-
ing dei nes the virtual computing environment for an application, including
the communication platforms, DVMs, and their coni gurations in a platform-
independent format. With VM modeling, the associated coni gurations
and management policies become an integrated part of a VE, which makes
swift redeployment and mobility possible. We address the instantiation and
deployment challenge by incarnating a VM from its model, which is signi-
i cantly more efi cient than the existing method of copying the predei ned
virtual machine image. As shown in
Figure 16.3
, users compose a VM with
the assistance of the VM Composition Service on the DVM Web portal when
preparing a job submission. To facilitate the user coni guration, predei ned
coni guration templates are collected and stored in the Knowledge Base
repository. The VE composition service may reference these templates when
composing VMs. After coni guration, a user submits the coni guration
together with its application and data to the VM Modeling Service. The Web
portal then generates a platform-independent VM model accordingly. When
a VM is scheduled to run on a physical platform, it is incarnated from its
model into a VM with the user application by the VM Incarnation and
Deployment Service. The realized virtual computing environment, as shown
in Figure 16.3, contains one or several VM hosting server(s) with interconnect-
ing networks mapped to physical resources. In this way, a DVM separates the
user domain and management domain and eases the virtual coni guration.
DVM mobility is designed to remove the dependency between the
virtualization layer and the physical layer, and hence to allow a DVM to
be scheduled and migrated dynamically. Existing solutions migrate a VM
via moving its image over the network [3,21]. This method is straightfor-
ward but it is very slow, due to the huge size of a system-level VM image,
and cannot support environment adaptation, which limits its applicability
in a grid environment.
The DVM approach is innovative. In its design, a VM image is decom-
posed into several functional units, which are saved on a high-speed virtual
Search WWH ::
Custom Search