[ovs-dev] RFC for database statistics interface.
blp at nicira.com
Wed Jun 9 02:31:35 UTC 2010
Thanks. I understand. I'll make sure that it works that way.
On Tue, Jun 8, 2010 at 6:43 PM, Jeremy Stribling <strib at nicira.com> wrote:
> The only use case I know is for the controller to poll the stats at a given
> interval, but monitor the statistics for delete events (so it can receive
> the last known value for the statistics). So as long as there's an
> understanding that the stats provider will update it one last time before
> deleting, what you propose will be fine.
> Ben Pfaff wrote:
>> I'm not sure that monitoring statistics columns is a great idea most of
>> the time. But supposing that a client does so, then I was planning to
>> just have it receive updates whenever something else causes the column
>> to be updated. That is, the presence or absence of a monitor doesn't in
>> itself cause columns to be updated.
>> I thought about a few other possibilities. One is that a monitor gets
>> updates as fast as they can be generated. I didn't seriously consider
>> that at all; I can't see any upsides. Another is that a monitor
>> specifies a maximum update frequency for statistics columns. There
>> might be some value in that, but the idea behind a statistics column is
>> that it changes very frequently, so there might not be any benefit over
>> polling to using a monitor for this purpose, since it would not reduce
>> the bandwidth required very much.
>> Do you have a use case that requires something smarter than what I have
>> in mind? There's no code yet, just documentation, so it's the perfect
>> time to change things.
>> On Tue, Jun 08, 2010 at 04:58:14PM -0700, Jeremy Stribling wrote:
>>> This looks ok to me, but how would this interact with monitoring?
>>> Will the client monitoring a particular statistics column only get
>>> an update after a client queries the stat? Is it the responsibility
>>> of the stats provider to update it every once in a while for the
>>> benefit of monitoring clients? Or maybe it doesn't make sense to
>>> monitor these statistics, and only query them occasionally.
>>> Ben Pfaff wrote:
>>>> I'm working on adding support for various kinds of statistics to the
>>>> OVS configuration database. I'm sending this patch out as a proposal
>>>> for the schema for statistics and for comments on how they would be
>>>> handled by the wire protocol for database access.
More information about the dev