You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@stratos.apache.org by Lahiru Sandaruwan <la...@wso2.com> on 2014/01/21 16:27:50 UTC
[Autoscaler] Scaling functionality to LB cartridge
Hi all,
I have been thinking on implementing $subject.
Currently Autoscaler receives requests in-flight stats grouped by cluster
and network partition from CEP.
I think we need requests in-flight stats grouped by network partition only
as well for $subject.
Because we have only one LB cluster per network partition.
Thanks.
--
--
Lahiru Sandaruwan
Software Engineer,
Platform Technologies,
WSO2 Inc., http://wso2.com
lean.enterprise.middleware
email: lahirus@wso2.com cell: (+94) 773 325 954
blog: http://lahiruwrites.blogspot.com/
twitter: http://twitter.com/lahirus
linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Imesh Gunaratne <im...@apache.org>.
I do agree on your point, may be a load balancer cartridge might not need
to be scaled similar to a application/framework cartridge.
However a load balancer might need to support high availability with
minimum of two instances in active-passive/active-active mode(s) and there
might be scenarios with extremely high incoming request loads which require
handful of instances.
Therefore I think the load balancer might need to have support to run in a
clustered environment. WDYT?
On Sun, Jan 26, 2014 at 8:14 AM, Afkham Azeez <af...@gmail.com> wrote:
>
>
>
> On Sun, Jan 26, 2014 at 2:38 PM, Afkham Azeez <af...@gmail.com> wrote:
>
>> I am still not convinced that autoscaling LB cartridges just like any
>> other app cartridges is the right thing to do. In fact, autoscaling data
>> cartridges the same was as app cartridges may not be the correct thing to
>> do.
>>
>
> s same was as app cartridges/same way as app cartridges
>
>
>> Registering/unregistering LB instances with DNS frequently when
>> autoscaling happens may not be the proper thing to do. The LB will never be
>> equivalent to a service like AWS ELB or Route 53, since Stratos LB
>> configuration & routing is an the application layer.
>>
>
> s/an the application layer/at the application layer
>
>
>> Treating the LB like any other normal cartridge is not the correct thing
>> to do, IMO.
>>
>> Azeez
>>
>>
>> On Sun, Jan 26, 2014 at 8:15 AM, Lakmal Warusawithana <la...@wso2.com>wrote:
>>
>>>
>>>
>>>
>>> On Sun, Jan 26, 2014 at 2:11 AM, Afkham Azeez <af...@gmail.com> wrote:
>>>
>>>> I think you need to think about how the autoscaled LB will receive
>>>> requests in the first place. Simply autoscaling the LB just because it
>>>> is possible is not that useful unless the story covers the part on how
>>>> those autoscaled LBs actually receive requests. I am wondering whether
>>>> autoscaling the LBs is even the correct thing to do.
>>>>
>>>
>>> When doing the horizontal scaling in cartridge layer, and if it fronted
>>> by single LBs then its going to bottleneck. Normally LB is the front face
>>> of the application, and request are routing to them by DNS. We can
>>> easily update DNS with newly created LBs. There are some use-cases I got to
>>> know with previous Stratos, they have make their LB as a cartridge and
>>> doing the same thing, since previous Stratos not provide LB as a cartridge.
>>>
>>>
>>>
>>>>
>>>>
>>>> On Tue, Jan 21, 2014 at 8:57 PM, Lahiru Sandaruwan <la...@wso2.com>wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> I have been thinking on implementing $subject.
>>>>>
>>>>> Currently Autoscaler receives requests in-flight stats grouped by
>>>>> cluster and network partition from CEP.
>>>>>
>>>>> I think we need requests in-flight stats grouped by network partition
>>>>> only as well for $subject.
>>>>> Because we have only one LB cluster per network partition.
>>>>>
>>>>> Thanks.
>>>>>
>>>>> --
>>>>> --
>>>>> Lahiru Sandaruwan
>>>>> Software Engineer,
>>>>> Platform Technologies,
>>>>> WSO2 Inc., http://wso2.com
>>>>> lean.enterprise.middleware
>>>>>
>>>>> email: lahirus@wso2.com cell: (+94) 773 325 954
>>>>> blog: http://lahiruwrites.blogspot.com/
>>>>> twitter: http://twitter.com/lahirus
>>>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> *Afkham Azeez*
>>>> Director of Architecture; WSO2, Inc.; http://wso2.com,
>>>> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>>>>
>>>> *email: **azeez@wso2.com* <az...@wso2.com>
>>>> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
>>>> *http://blog.afkham.org* <http://blog.afkham.org>
>>>> * twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
>>>> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
>>>> <http://lk.linkedin.com/in/afkhamazeez>*
>>>>
>>>> *Lean . Enterprise . Middleware*
>>>>
>>>>
>>>
>>>
>>> --
>>> Lakmal Warusawithana
>>> Software Architect; WSO2 Inc.
>>> Mobile : +94714289692
>>> Blog : http://lakmalsview.blogspot.com/
>>>
>>>
>>
>>
>> --
>> *Afkham Azeez*
>> Director of Architecture; WSO2, Inc.; http://wso2.com,
>> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>>
>> *email: **azeez@wso2.com* <az...@wso2.com>
>> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
>> *http://blog.afkham.org* <http://blog.afkham.org>
>> *twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
>> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
>> <http://lk.linkedin.com/in/afkhamazeez>*
>>
>> *Lean . Enterprise . Middleware*
>>
>>
>
>
> --
> *Afkham Azeez*
> Director of Architecture; WSO2, Inc.; http://wso2.com,
> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>
> *email: **azeez@wso2.com* <az...@wso2.com>
> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
> *http://blog.afkham.org* <http://blog.afkham.org>
> *twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
> <http://lk.linkedin.com/in/afkhamazeez>*
>
> *Lean . Enterprise . Middleware*
>
>
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Afkham Azeez <af...@gmail.com>.
On Sun, Jan 26, 2014 at 2:38 PM, Afkham Azeez <af...@gmail.com> wrote:
> I am still not convinced that autoscaling LB cartridges just like any
> other app cartridges is the right thing to do. In fact, autoscaling data
> cartridges the same was as app cartridges may not be the correct thing to
> do.
>
s same was as app cartridges/same way as app cartridges
> Registering/unregistering LB instances with DNS frequently when
> autoscaling happens may not be the proper thing to do. The LB will never be
> equivalent to a service like AWS ELB or Route 53, since Stratos LB
> configuration & routing is an the application layer.
>
s/an the application layer/at the application layer
> Treating the LB like any other normal cartridge is not the correct thing
> to do, IMO.
>
> Azeez
>
>
> On Sun, Jan 26, 2014 at 8:15 AM, Lakmal Warusawithana <la...@wso2.com>wrote:
>
>>
>>
>>
>> On Sun, Jan 26, 2014 at 2:11 AM, Afkham Azeez <af...@gmail.com> wrote:
>>
>>> I think you need to think about how the autoscaled LB will receive
>>> requests in the first place. Simply autoscaling the LB just because it
>>> is possible is not that useful unless the story covers the part on how
>>> those autoscaled LBs actually receive requests. I am wondering whether
>>> autoscaling the LBs is even the correct thing to do.
>>>
>>
>> When doing the horizontal scaling in cartridge layer, and if it fronted
>> by single LBs then its going to bottleneck. Normally LB is the front face
>> of the application, and request are routing to them by DNS. We can
>> easily update DNS with newly created LBs. There are some use-cases I got to
>> know with previous Stratos, they have make their LB as a cartridge and
>> doing the same thing, since previous Stratos not provide LB as a cartridge.
>>
>>
>>
>>>
>>>
>>> On Tue, Jan 21, 2014 at 8:57 PM, Lahiru Sandaruwan <la...@wso2.com>wrote:
>>>
>>>> Hi all,
>>>>
>>>> I have been thinking on implementing $subject.
>>>>
>>>> Currently Autoscaler receives requests in-flight stats grouped by
>>>> cluster and network partition from CEP.
>>>>
>>>> I think we need requests in-flight stats grouped by network partition
>>>> only as well for $subject.
>>>> Because we have only one LB cluster per network partition.
>>>>
>>>> Thanks.
>>>>
>>>> --
>>>> --
>>>> Lahiru Sandaruwan
>>>> Software Engineer,
>>>> Platform Technologies,
>>>> WSO2 Inc., http://wso2.com
>>>> lean.enterprise.middleware
>>>>
>>>> email: lahirus@wso2.com cell: (+94) 773 325 954
>>>> blog: http://lahiruwrites.blogspot.com/
>>>> twitter: http://twitter.com/lahirus
>>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>>>>
>>>>
>>>
>>>
>>> --
>>> *Afkham Azeez*
>>> Director of Architecture; WSO2, Inc.; http://wso2.com,
>>> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>>>
>>> *email: **azeez@wso2.com* <az...@wso2.com>
>>> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
>>> *http://blog.afkham.org* <http://blog.afkham.org>
>>> * twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
>>> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
>>> <http://lk.linkedin.com/in/afkhamazeez>*
>>>
>>> *Lean . Enterprise . Middleware*
>>>
>>>
>>
>>
>> --
>> Lakmal Warusawithana
>> Software Architect; WSO2 Inc.
>> Mobile : +94714289692
>> Blog : http://lakmalsview.blogspot.com/
>>
>>
>
>
> --
> *Afkham Azeez*
> Director of Architecture; WSO2, Inc.; http://wso2.com,
> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>
> *email: **azeez@wso2.com* <az...@wso2.com>
> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
> *http://blog.afkham.org* <http://blog.afkham.org>
> *twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
> <http://lk.linkedin.com/in/afkhamazeez>*
>
> *Lean . Enterprise . Middleware*
>
>
--
*Afkham Azeez*
Director of Architecture; WSO2, Inc.; http://wso2.com,
*Member; Apache Software Foundation;
**http://www.apache.org/*<http://www.apache.org/>
*email: **azeez@wso2.com* <az...@wso2.com>
* cell: +94 77 3320919 blog: **http://blog.afkham.org*<http://blog.afkham.org>
*twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
* linked-in: **http://lk.linkedin.com/in/afkhamazeez
<http://lk.linkedin.com/in/afkhamazeez>*
*Lean . Enterprise . Middleware*
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Afkham Azeez <af...@gmail.com>.
I am still not convinced that autoscaling LB cartridges just like any other
app cartridges is the right thing to do. In fact, autoscaling data
cartridges the same was as app cartridges may not be the correct thing to
do. Registering/unregistering LB instances with DNS frequently when
autoscaling happens may not be the proper thing to do. The LB will never be
equivalent to a service like AWS ELB or Route 53, since Stratos LB
configuration & routing is an the application layer. Treating the LB like
any other normal cartridge is not the correct thing to do, IMO.
Azeez
On Sun, Jan 26, 2014 at 8:15 AM, Lakmal Warusawithana <la...@wso2.com>wrote:
>
>
>
> On Sun, Jan 26, 2014 at 2:11 AM, Afkham Azeez <af...@gmail.com> wrote:
>
>> I think you need to think about how the autoscaled LB will receive
>> requests in the first place. Simply autoscaling the LB just because it
>> is possible is not that useful unless the story covers the part on how
>> those autoscaled LBs actually receive requests. I am wondering whether
>> autoscaling the LBs is even the correct thing to do.
>>
>
> When doing the horizontal scaling in cartridge layer, and if it fronted by
> single LBs then its going to bottleneck. Normally LB is the front face of
> the application, and request are routing to them by DNS. We can easily
> update DNS with newly created LBs. There are some use-cases I got to know
> with previous Stratos, they have make their LB as a cartridge and doing the
> same thing, since previous Stratos not provide LB as a cartridge.
>
>
>
>>
>>
>> On Tue, Jan 21, 2014 at 8:57 PM, Lahiru Sandaruwan <la...@wso2.com>wrote:
>>
>>> Hi all,
>>>
>>> I have been thinking on implementing $subject.
>>>
>>> Currently Autoscaler receives requests in-flight stats grouped by
>>> cluster and network partition from CEP.
>>>
>>> I think we need requests in-flight stats grouped by network partition
>>> only as well for $subject.
>>> Because we have only one LB cluster per network partition.
>>>
>>> Thanks.
>>>
>>> --
>>> --
>>> Lahiru Sandaruwan
>>> Software Engineer,
>>> Platform Technologies,
>>> WSO2 Inc., http://wso2.com
>>> lean.enterprise.middleware
>>>
>>> email: lahirus@wso2.com cell: (+94) 773 325 954
>>> blog: http://lahiruwrites.blogspot.com/
>>> twitter: http://twitter.com/lahirus
>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>>>
>>>
>>
>>
>> --
>> *Afkham Azeez*
>> Director of Architecture; WSO2, Inc.; http://wso2.com,
>> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>>
>> *email: **azeez@wso2.com* <az...@wso2.com>
>> * cell: +94 77 3320919 <%2B94%2077%203320919> blog: *
>> *http://blog.afkham.org* <http://blog.afkham.org>
>> *twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
>> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
>> <http://lk.linkedin.com/in/afkhamazeez>*
>>
>> *Lean . Enterprise . Middleware*
>>
>>
>
>
> --
> Lakmal Warusawithana
> Software Architect; WSO2 Inc.
> Mobile : +94714289692
> Blog : http://lakmalsview.blogspot.com/
>
>
--
*Afkham Azeez*
Director of Architecture; WSO2, Inc.; http://wso2.com,
*Member; Apache Software Foundation;
**http://www.apache.org/*<http://www.apache.org/>
*email: **azeez@wso2.com* <az...@wso2.com>
* cell: +94 77 3320919 blog: **http://blog.afkham.org*<http://blog.afkham.org>
*twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
* linked-in: **http://lk.linkedin.com/in/afkhamazeez
<http://lk.linkedin.com/in/afkhamazeez>*
*Lean . Enterprise . Middleware*
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Lakmal Warusawithana <la...@wso2.com>.
On Sun, Jan 26, 2014 at 2:11 AM, Afkham Azeez <af...@gmail.com> wrote:
> I think you need to think about how the autoscaled LB will receive
> requests in the first place. Simply autoscaling the LB just because it is
> possible is not that useful unless the story covers the part on how those
> autoscaled LBs actually receive requests. I am wondering whether
> autoscaling the LBs is even the correct thing to do.
>
When doing the horizontal scaling in cartridge layer, and if it fronted by
single LBs then its going to bottleneck. Normally LB is the front face of
the application, and request are routing to them by DNS. We can easily
update DNS with newly created LBs. There are some use-cases I got to know
with previous Stratos, they have make their LB as a cartridge and doing the
same thing, since previous Stratos not provide LB as a cartridge.
>
>
> On Tue, Jan 21, 2014 at 8:57 PM, Lahiru Sandaruwan <la...@wso2.com>wrote:
>
>> Hi all,
>>
>> I have been thinking on implementing $subject.
>>
>> Currently Autoscaler receives requests in-flight stats grouped by cluster
>> and network partition from CEP.
>>
>> I think we need requests in-flight stats grouped by network partition
>> only as well for $subject.
>> Because we have only one LB cluster per network partition.
>>
>> Thanks.
>>
>> --
>> --
>> Lahiru Sandaruwan
>> Software Engineer,
>> Platform Technologies,
>> WSO2 Inc., http://wso2.com
>> lean.enterprise.middleware
>>
>> email: lahirus@wso2.com cell: (+94) 773 325 954
>> blog: http://lahiruwrites.blogspot.com/
>> twitter: http://twitter.com/lahirus
>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>>
>>
>
>
> --
> *Afkham Azeez*
> Director of Architecture; WSO2, Inc.; http://wso2.com,
> *Member; Apache Software Foundation; **http://www.apache.org/*<http://www.apache.org/>
>
> *email: **azeez@wso2.com* <az...@wso2.com>
> * cell: +94 77 3320919 blog: **http://blog.afkham.org*<http://blog.afkham.org>
> *twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
> * linked-in: **http://lk.linkedin.com/in/afkhamazeez
> <http://lk.linkedin.com/in/afkhamazeez>*
>
> *Lean . Enterprise . Middleware*
>
>
--
Lakmal Warusawithana
Software Architect; WSO2 Inc.
Mobile : +94714289692
Blog : http://lakmalsview.blogspot.com/
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Afkham Azeez <af...@gmail.com>.
I think you need to think about how the autoscaled LB will receive requests
in the first place. Simply autoscaling the LB just because it is possible
is not that useful unless the story covers the part on how those autoscaled
LBs actually receive requests. I am wondering whether autoscaling the LBs
is even the correct thing to do.
On Tue, Jan 21, 2014 at 8:57 PM, Lahiru Sandaruwan <la...@wso2.com> wrote:
> Hi all,
>
> I have been thinking on implementing $subject.
>
> Currently Autoscaler receives requests in-flight stats grouped by cluster
> and network partition from CEP.
>
> I think we need requests in-flight stats grouped by network partition
> only as well for $subject.
> Because we have only one LB cluster per network partition.
>
> Thanks.
>
> --
> --
> Lahiru Sandaruwan
> Software Engineer,
> Platform Technologies,
> WSO2 Inc., http://wso2.com
> lean.enterprise.middleware
>
> email: lahirus@wso2.com cell: (+94) 773 325 954
> blog: http://lahiruwrites.blogspot.com/
> twitter: http://twitter.com/lahirus
> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>
>
--
*Afkham Azeez*
Director of Architecture; WSO2, Inc.; http://wso2.com,
*Member; Apache Software Foundation;
**http://www.apache.org/*<http://www.apache.org/>
*email: **azeez@wso2.com* <az...@wso2.com>
* cell: +94 77 3320919 blog: **http://blog.afkham.org*<http://blog.afkham.org>
*twitter: **http://twitter.com/afkham_azeez*<http://twitter.com/afkham_azeez>
* linked-in: **http://lk.linkedin.com/in/afkhamazeez
<http://lk.linkedin.com/in/afkhamazeez>*
*Lean . Enterprise . Middleware*
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Lahiru Sandaruwan <la...@wso2.com>.
Yes Imash. Great. We need that part first.
Thanks.
On Thu, Jan 23, 2014 at 5:59 PM, Imesh Gunaratne <im...@apache.org> wrote:
> Hi Lahiru,
>
> Thanks for looking into this. I think first we need to implement load
> balancer state replication functionality to make it possible to run
> multiple instances of the load balancer in a cluster.
>
> I will start on this in parallel.
>
> Thanks
> Imesh
>
>
> On Tue, Jan 21, 2014 at 10:27 AM, Lahiru Sandaruwan <la...@wso2.com>wrote:
>
>> Hi all,
>>
>> I have been thinking on implementing $subject.
>>
>> Currently Autoscaler receives requests in-flight stats grouped by cluster
>> and network partition from CEP.
>>
>> I think we need requests in-flight stats grouped by network partition
>> only as well for $subject.
>> Because we have only one LB cluster per network partition.
>>
>> Thanks.
>>
>> --
>> --
>> Lahiru Sandaruwan
>> Software Engineer,
>> Platform Technologies,
>> WSO2 Inc., http://wso2.com
>> lean.enterprise.middleware
>>
>> email: lahirus@wso2.com cell: (+94) 773 325 954
>> blog: http://lahiruwrites.blogspot.com/
>> twitter: http://twitter.com/lahirus
>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>>
>>
>
--
--
Lahiru Sandaruwan
Software Engineer,
Platform Technologies,
WSO2 Inc., http://wso2.com
lean.enterprise.middleware
email: lahirus@wso2.com cell: (+94) 773 325 954
blog: http://lahiruwrites.blogspot.com/
twitter: http://twitter.com/lahirus
linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
Re: [Autoscaler] Scaling functionality to LB cartridge
Posted by Imesh Gunaratne <im...@apache.org>.
Hi Lahiru,
Thanks for looking into this. I think first we need to implement load
balancer state replication functionality to make it possible to run
multiple instances of the load balancer in a cluster.
I will start on this in parallel.
Thanks
Imesh
On Tue, Jan 21, 2014 at 10:27 AM, Lahiru Sandaruwan <la...@wso2.com>wrote:
> Hi all,
>
> I have been thinking on implementing $subject.
>
> Currently Autoscaler receives requests in-flight stats grouped by cluster
> and network partition from CEP.
>
> I think we need requests in-flight stats grouped by network partition
> only as well for $subject.
> Because we have only one LB cluster per network partition.
>
> Thanks.
>
> --
> --
> Lahiru Sandaruwan
> Software Engineer,
> Platform Technologies,
> WSO2 Inc., http://wso2.com
> lean.enterprise.middleware
>
> email: lahirus@wso2.com cell: (+94) 773 325 954
> blog: http://lahiruwrites.blogspot.com/
> twitter: http://twitter.com/lahirus
> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
>
>