Method and Apparatus to introduce a new FC filter in the PRS for providing intelligent resource scheduling for NPIV device connections.
Publication Date: 2017-Aug-04
The IP.com Prior Art Database
Method and apparatus to introduce a new FC filter in the scheduler for providing intelligent resource scheduling for NPIV device connections. Abstract: Disclosed is a method which will improve the intelligence of the scheduler while making a virtual machine placement decision during deployment and migration on a host by introducing a new filter to track the available connections on FC ports of every host. Proposed Solution: In an NPIV cloud world, one of the important parameters to be considered during placement is to check the available connections on the fibre channel [ FC] port of a host which is currently missing in the scheduler [Scheduler: provides dynamic resource management for OpenStack cloud environments to optimize business agility and reduce cost of cloud ownership.]metrics because of which the virtual machines get scheduled to a host whose FC port has exhausted all the available connections thereby creating unusable virtual machines that fail to boot up. Adding intelligence to the scheduler by introducing a new dynamic metric namely FCportFilter to evaluate the free connections on an FC port for every host during the time of making a scheduling decision for a virtual machine's placement will resolve this problem. When virtual machines are scheduled to a host which has no free connections on the FC port, such virtual machines fail to acquire an FLOGI [ Fabric login: A login server is a service running inside each fabric switch. It is responsible for providing a unique FCID (fibre channel id) to each device connected to the switch, just like the DHCP provides each network host with a unique IP address. The FCID gets assigned to the virtual FC adapter which accesses the storage volume and finally boots up the virtual machine.] When the virtual machine does not acquire an FLOGI, it will be unable to connect to the underlying storage volume, the boot process fails and the virtual machine becomes unusable. All the resources allocated to the virtual machine are also unutilized. In a cloud environment, when virtual machines are deployed in huge numbers, there is a resource-aware and workload-smart solution called as scheduler which gets exposed to a flexible resource selection capability and a rich set of dynamic and static metrics reflecting the state of hosts and hypervisors using which the cloud admins and users can make better use of infrastructure resources. Hints can be provided to the scheduler in the form of resource requirement expressions and the scheduler will seek to find the optimal resources for application workloads to maximize application performance and availability. Illustrating the problem with a cloud virtualization software as an example. As shown in Figure 1, all the 64 connections on fcs0 on VIOS_1 and VIOS_2 have exhausted. But the scheduler is unaware of this situation and it still considers both the hosts as valid targets for deploys. Deploys will get triggered to these hosts, but those virtual machines will ne...