Dashboard graphs require instance hours graph

[Tim Hoffman]

Hi

To help us get a better grasp on hour instance hours consumption we really need an graph of

instance hours used per hour.  The current graph of instances doesn't really convey that information, 

At the moment there is no indication of instance hours use other than in the billing stats 

which is running behind and has no granularity. When trying to optimise our instance hours use, this would

be very helpful. 

Or is there some other indicator that exists but I haven't noticed ?

It would seem remiss to charge based on the criteria but not actually provide any direct measurement of it.

Thanks 

Tim

[Gregory D'alesandre]

Hi Tim, 

The drop down on the charts page has "Instances" as the 2nd to last item in the menu.  Is that what you are looking for?

Hope that helps,

Greg

--
You received this message because you are subscribed to the Google Groups "Google App Engine" group.
To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/gIE6ynFMNLsJ.
To post to this group, send email to google-a...@googlegroups.com.
To unsubscribe from this group, send email to google-appengi...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.

[Tim Hoffman]

Hi Greg

You didn't really read my post. I explicitly said the graph of instances doesn't really help.

How do you guage you instance hours accumulated for the hour from the instances graph that is currently there.

Maybe a graph isn't the right tool, Can you tell me how many instance hours and therefore dollars I have accumulated for each hour, half hour, 10 min (name you interval) from this graph ?

Yet we are going to be paying for instance hours.  Yet can't get a breakdown of actual cost as we go.  Also note the oddness of the active line in the is graph.  But thats another store.

Rgds

Tim

[Barry Hunter]

On Fri, Sep 2, 2011 at 1:35 PM, Tim Hoffman <zute...@gmail.com> wrote:

Hi Greg

You didn't really read my post. I explicitly said the graph of instances doesn't really help.

How do you guage you instance hours accumulated for the hour from the instances graph that is currently there.

Maybe a graph isn't the right tool, Can you tell me how many instance hours and therefore dollars I have accumulated for each hour, half hour, 10 min (name you interval) from this graph ?

The instance hours is the AREA under the line. (because the graph is instances and time plotted together)

So take a column say a hour wide. And then the instance hours used in that hour, is the height of the bar covered by that column. You would have to approximate the 'average' top of the bar. 

So on your graph, between -7 and -6 hours, is about 7.75 instance hours. 

Would be easier to work out if there where more vertical gridlines. 

(although this doesnt account for the 15 minute startup fee. But assuming there isnt much churn in your instances, can see roughly how many intances would of started (by the upslopes) )

 

Yet we are going to be paying for instance hours.  Yet can't get a breakdown of actual cost as we go.  Also note the oddness of the active line in the is graph.  But thats another store.

Rgds

Tim

--

You received this message because you are subscribed to the Google Groups "Google App Engine" group.

To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/oEWS7b4pJjgJ.

[Tim Hoffman]

HI Barry

I realize they area under the graph is number of instances, however in my case there is in fact a lot of instance churn (especially when I drop off always on)

Like I said later maybe graphs aren't the best way to provide this information.

I do want to see the rate of instance creation, vs rate of removal, 

Pretty hard to build up real cost trends from graphs like these, underlying data would actually be more useful.  Takes a lot more effort

than a quick glance to work out my cost for the last hour.  In addition this graph doesn't distinguish between backend, frontend and 

precommitted hours, all of which can have different charges.

Hey and I am lazy, and this is one of the primary metrics we will be charged on.

T

[Gregory D'alesandre]

Hi Tim,

It's a fair question.  These graphs only show frontend instances.  Precommitted Instance Hours don't actually operate differently, they are only billed differently, so we are not planning to show them on this graph but rather on your bill.  All of that being said, this graph should be more useful and we'll see what we can do to improve it.

Thanks for the question,

Greg

T

--

You received this message because you are subscribed to the Google Groups "Google App Engine" group.

To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/lSJVu1WboWYJ.

[Tim Hoffman]

HI Greg

Yep the graph might not be that useful, however billing stats are running at about 3 days behind, so between late billing stats and a less

than useful graph in terms of estimating your actual costs the round trip time on making real changes and see the true impact on real costs 

is pretty long - like 3+ days at the moment.

A graph of instances is hard to relate to real costs as we go because we can;t relate instances directly to costs given the different charge rates.. 

So hard realtime stats are a pretty hot button at the moment ;-)

Rgds

T

[Tim Hoffman]

Hi

I have been thinking further about what I think the bare minimum the graphs should be showing.

It really needs instance hours cost on 15 min increments (or less) or if  that is not possible at least instance hours,

number of instances isn't the metric we are going to be charged on. 

For bursty traffic number of instances gives you absolutely no idea what you are really going to be charged, and pretty much impossible

looking at the graphs  and given the statement  that we are going to be charged max idle instances + active instances  (ie never charged 

for idle instances > max idle instances) the total number of instances is pretty irrelevant.

Getting import times down and more consistancy in startup times could mean getting rid of any resident idle instances altogether.  

But thats another topic entirely ;-)

Rgds

T