Re: app auto-scaling in OSS CF contribution

john mcteague <john.mcteague@...>

One of the areas of autoscaling we are looking at is in application led
autoscaling, by which I do not mean the app's themselves calling an API to
scale up or down, but having an agreed contract/API between app and
autoscaler by which the application can flag the need to scale up or down
based on app specific metrics, not just metrics that are visible via the
standard firehose/CC API. The app would not call an API, it may just expose
an API endpoint that the autoscaler would poll.

An example may be, if the depth of a message queue exceeds a certain limit,
we need more instances of that app to consume all messages within a given
SLA. We should be able to automate this in an autoscaler.


On Tue, Sep 15, 2015 at 4:42 PM, Klevenz, Stephan <stephan.klevenz(a)>


A contribution to the incubator about that feature is very interesting and
will get my attention.


Von: Siva Balan
Antworten an: "Discussions about Cloud Foundry projects and the system
Datum: Dienstag, 15. September 2015 17:20
An: "Discussions about Cloud Foundry projects and the system overall."
Betreff: [cf-dev] Re: Re: Re: app auto-scaling in OSS CF contribution

+1 on this feature. We at GE would be very interested in this feature as
well. We would very much like to collaborate on this feature.


On Tue, Sep 15, 2015 at 7:59 AM, Guillaume Berche <bercheg(a)>

Hi Dies,

Thanks for your future sharing of your work on an opensource autoscaler

We would be interested at Orange to use the autoscaler, and potentially
try to plug into it the autosleep service [1] we're working on, so that a
minimum instance count the autoscaler could set could be equal to zero.

To me, one important missing aspect to make use of an autoscaling service
to handle changes in workloads, is to effectively handle newly instanciated
cold instances (i.e. whose lazy initialized caches are not yet warm, and
which would degrade user perceived experience if given immediately a fixed
% of the traffic):
- either to have support for autoscaling service to send warmup http
requests to newly created instances (similar to GAE warmup support)
- have the gorouter support a traffic ramping setting, so that a cold
instance is slowly getting traffic when entering rotation.




On Tue, Sep 15, 2015 at 2:30 AM, ronak banka <ronakbanka.cse(a)>

Hi Dies,

App auto-scaling is much needed feature for CF OSS , lot of users want
to use this functionality .

Once on the incubator, roadmap can be discussed. Hope to to see it soon
on cf incubator .

Ronak Banka
Rakuten, Inc.

On Tue, Sep 15, 2015 at 9:00 AM, Koper, Dies <diesk(a)>


At Fujitsu we’re developing app auto-scaling and are considering to
propose moving it to the cf incubator.

Before we start open-sourcing it, I wanted to ask if there is any
interest for this in the community, possibly even others working on or
considering to work on one who’d be interested to collaborate/align with us?

We’re looking at providing basic support for scaling up/down based on
metrics like CPU, request count, and a service broker to enable it for your

We can share a detailed functional description for review in a few

Depending on priorities, interest and resources available we may add
functionality like sending an email notification in addition to/instead of
scaling, or scale based on other metrics (including app generated custom

Either way, we want to make these things pluggable to allow people to
integrate it with their own (closed-source) monitoring agents or custom

I feel every PaaS comes with free app auto-scaling functionality (PCF,
Bluemix, OpenShift, AWS, …) so OSS CF deserves one too.

I have discussed this plan with Pivotal and they have encouraged me to
send this email to the list.

Please let me know if you have any questions.


Dies Koper



Join { to automatically receive all group messages.