[
https://issues.apache.org/jira/browse/CLOUDSTACK-235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13554947#comment-13554947
]
Radhika Nair commented on CLOUDSTACK-235:
-----------------------------------------
https://reviews.apache.org/r/8970/
> Network rate can be set in 2 places. Clarify docs on how this works.
> --------------------------------------------------------------------
>
> Key: CLOUDSTACK-235
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-235
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: Doc
> Affects Versions: 4.0.0
> Reporter: Jessica Tomechak
> Assignee: Radhika Nair
> Priority: Minor
> Fix For: 4.1.0
>
> Attachments: network-rate.html
>
>
> What is the purpose of the two Network Rates. There is one in Compute
> Offerings and one in Network Offerings. How does each apply in basic &
> advanced networking? (Michael Simos)
> (Kirk Kosinski:)
> With vSphere, the actual limits vary depending on:
> 1. Where they are configured (compute and/or network offering) 2. The
> network type (shared or isolated) 3. The traffic direction (ingress or
> egress)
>
> I'd assume that a basic zone would work like a shared network in an
> advanced zone, but if not, add that to the list above. However, it
> may function differently in XenServer, so hypervisor might also need
> to be on the list (and even if XenServer and vSphere function the
> same, KVM doesn't support limits at all). Also, it is probably different in
> vSphere with Nexus 1000V since (I think) ingress traffic can be limited (a
> regular dvSwitch can limit ingress/egress, and I think the Nexus 1000V is
> considered a dvSwitch... but I only tested with regular vSwitches, which can
> only limit egress)... so...vSwitch type may need to be on that list.
> Network Rate can be configured on either the Network
> Offering or Compute Offering, on both of them simultaneously, or on
> neither of them. The resulting behavior in vSphere is complicated. However, I
> will try to explain.
>
> The Network Rate for a Network Offering used by a particular network
> in CloudStack will be used for the traffic shaping policy of a port
> group for that network (i.e. a particular subnet/VLAN on the actual network).
> Virtual routers for that network will connect to this port group, and by
> default instances in that network will connect to this port group.
> However, if an instance is deployed with a Compute Offering with a
> Network Rate, this rate will be used for the traffic shaping policy of
> another port group for the network, and instances using the offering will be
> connected to this port group instead.
>
> Traffic shaping on standard port groups in vSphere only applies to
> egress traffic and the net effect depends on the type of network in
> CloudStack. For shared networks, ingress traffic is unlimited as far
> as CloudStack is concerned, and egress traffic is limited to the rate
> that applies to the port group used by the instance (if any). If the
> Compute Offering has a Network Rate configured, this rate will apply
> to egress traffic, otherwise the Network Rate of the Network Offering will
> apply. For isolated networks, the Network Rate for the Network Offering (if
> any) will effectively apply to ingress traffic (since it applies to egress
> traffic from the virtual router to the instance), and egress traffic is
> limited to the rate that applies to the port group used by the instance (if
> any), similar to shared networks.
>
> So for example:
> Network Rate of Network Offering = 10 Mb/s
> Network Rate of Compute Offering = 200 Mb/s
>
> In a shared network, ingress traffic will not be limited as far as
> CloudStack is concerned, while egress traffic will be limited to 200 Mb/s. In
> an isolated network, ingress traffic will be limited to 10 Mb/s and egress to
> 200 Mb/s.
> (Kirk Kosinski)
> See: http://docs.cloudstack.org/Knowledge_Base/Network_Throttling. We have
> confirmed the current code behaves as documented here (Murali Reddy)
> It is different in vSphere with Nexus 1000V since ingress traffic can be
> limited, as well as egress traffic. (Sateesh Chodapuneedi)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
