Public bug reported: Problem description: ******************** In a multi-tenant environment, some VMs could consume the available bandwidth and cause network congestion for those that share the same network resources. To improve the quality of service, openstack implements a service to limit the bandwidth of VM ports and classify the traffic using DSCP field. However, there is still a lack of a mechanism policing specific VMs that consume a lot of resources and generate network congestion.
API effects: ************ By adding a new rule in the QoS extension [1] of neutron we could detect and react dynamically to the congestion. This rule will police VMs causing congestion by limiting the allowed bandwidth in their ports. Proposal: ********* Our proposal is to use Explicit Congestion Notification (ECN) [2] bits as marking mechanism. So, the network near congestion will mark packets as Congestion Experienced (CE). Neutron will detect these packets and starts monitoring the behavior of the VM causing the congestion. When the congestion exceeding a threshold, neutron will react by policing this VM using the bandwidth limitation rule of QoS extension. Neutron will keep track of the congestion rate and will free the policed VMs when the congestion is over. References: *********** [1] QoS https://docs.openstack.org/ocata/networking-guide/config-qos.html [2] ECN in IP https://tools.ietf.org/html/rfc3168 ** Affects: neutron Importance: Undecided Status: New ** Tags: qos -- You received this bug notification because you are a member of Yahoo! Engineering Team, which is subscribed to neutron. https://bugs.launchpad.net/bugs/1708460 Title: [RFE] Reaction to network congestion for qos Status in neutron: New Bug description: Problem description: ******************** In a multi-tenant environment, some VMs could consume the available bandwidth and cause network congestion for those that share the same network resources. To improve the quality of service, openstack implements a service to limit the bandwidth of VM ports and classify the traffic using DSCP field. However, there is still a lack of a mechanism policing specific VMs that consume a lot of resources and generate network congestion. API effects: ************ By adding a new rule in the QoS extension [1] of neutron we could detect and react dynamically to the congestion. This rule will police VMs causing congestion by limiting the allowed bandwidth in their ports. Proposal: ********* Our proposal is to use Explicit Congestion Notification (ECN) [2] bits as marking mechanism. So, the network near congestion will mark packets as Congestion Experienced (CE). Neutron will detect these packets and starts monitoring the behavior of the VM causing the congestion. When the congestion exceeding a threshold, neutron will react by policing this VM using the bandwidth limitation rule of QoS extension. Neutron will keep track of the congestion rate and will free the policed VMs when the congestion is over. References: *********** [1] QoS https://docs.openstack.org/ocata/networking-guide/config-qos.html [2] ECN in IP https://tools.ietf.org/html/rfc3168 To manage notifications about this bug go to: https://bugs.launchpad.net/neutron/+bug/1708460/+subscriptions -- Mailing list: https://launchpad.net/~yahoo-eng-team Post to : yahoo-eng-team@lists.launchpad.net Unsubscribe : https://launchpad.net/~yahoo-eng-team More help : https://help.launchpad.net/ListHelp
