Re: [mongrel2] A killer app for Mongrel2
- From:
- David Kantowitz
- Date:
- 2011-12-31 @ 21:49
Hey Lee,
I think there is more to be done with PaaS. I do like the picloud approach
of 'cloud-izing' at the function level and charging by the millisecond
instead of cloud-izing at the virtual-server level. I think the picloud
approach adds a layer of actual value beyond what most PaaS companies do --
which, imho, is some flavor of a deployment tool.
I'm envisioning something where you can run your entire php/python/whatever
web application on the BarCloud and only pay for the time where your
functions are actually executing. All the other stuff about webservers,
server instance, memory, disk, etc are all handled for you. If you get 1
hit a day .. no problem; if you get 10M hits a day, also no problem.
Internally the thing would basically be a clever "cloud computation"
scheduler that spun up EC2 instances as needed based on the probability
distributions of the loads to minimize cost for a given response time [a
problem mathematicallhy very similar to what I've been working on for my
stock trading systems].
This seems like it would add a layer of actual value by transforming the
interface level that a developer uses to interface with the cloud. Further
it allows a platform for some real cross-cloud brokering.
David
PS. I've been using the Python(x,y) distribution lately. It's been working
really well for me (better then the activestate distribution which used to
be my python setup of choice).
On Sat, Dec 31, 2011 at 10:46 AM, Kenneth Keiter <ken@kenkeiter.com> wrote:
> I think that a genuine 'killer-app' for Mongrel2 will be one which takes
> advantage of what Mongrel2 specializes in. There are several situations in
> which I would not use it because, frankly, there are better tools for the
> job. But the best servers are opinionated in their design -- they tackle
> one set of tasks incredibly well -- and Mongrel2 is no exception.
>
> Although *my* killer app is not open source, I take advantage of
> Mongrel2's use of ZeroMQ (and frequently its ability to address many
> clients with a single response) to flexibly scale client-facing
> applications that I host. To do so, I've been developing my own PaaS
> solution, which I call Tapestry. Here are some details:
>
> Tapestry's primary purpose is to allow simplified distribution and
> automatic scaling of Ruby, Python, and Lua web applications/services on EC2
> infrastructure. It uses a master/slave architecture in which a dedicated
> EC2 machine with elastic IP is setup as a master, and runs Mongrel2. It
> also acts as a hub with which slaves will communicate (via MessagePack RPC
> spec over ZeroMQ). Slaves serve to support the modular web applications,
> and are deployed automatically as new EC2 instances by the master. They
> pull application code from a repo (using signed tags for verification), and
> a bundled setup descriptor allows the application to reconfigure the slave
> to support it. Here's the primary place where Mongrel2 comes in: the
> Tapestry master reconfigures Mongrel2 to serve new applications
> automatically, and the application instances running on separate slaves
> connect back to the master's Mongrel2 instances via ZeroMQ over TCP to
> scale the application horizontally. Multiple masters are allowed, and new
> masters are advertised over mDNS. Voila!
> Auto-horizontally/vertically-scaling (where appropriate), easily
> distributed and updated web applications. Pretty simple.
>
> Anyway, my point is that there's a proper tool for the job in every
> circumstance. If you're talking killer apps, I think the best choice is
> going to be one that really illustrates how Mongrel2 gets out of the way
> and makes serving/scaling incredible web applications into a clean and
> simple process. We need an idea that illustrates:
>
> * ease of configuration (and reconfiguration on the fly);
> * topological/architectural flexibility afforded by Mongrel2's use of
> ZeroMQ;
> * speed/ability to support vast numbers of clients;
> * naught to up-and-running ease;
> * ease of converting existing applications.
>
> Thoughts?
>
> Happy new year to everybody.
>
> Cheers,
> Ken Keiter
>
> On Saturday, December 31, 2011 at 10:17 AM, Mike Laidlaw wrote:
>
> Not sure if this exactly qualifies as a killer app, but I've been
> thinking about this problem: my sister-in-law and I were discussing the
> hours of personal digital video that we have collected (kids-family
> stuff) and what to do with it all.
>
> I was thinking it would be great if I could stick it all on a shared
> server and then be able to access it with a youtube like interface from
> any web-browser anywhere. The idea is we have total control over all of
> the video, but we could use it just like youtube. Maybe there is
> something like that out there already? Any thoughts?
>
> Happy New Year to all!
>
> Mike
>
>
> On 10/12/2011 3:26 AM, Loic d'Anterroches wrote:
>
> Hello,
>
> no, I am not announcing a killer app for Mongrel2 which will make
> everybody switch from Apache/Nginx to Mongrel2.
>
> I just wonder, what could be a killer app which could really take
> benefit of the Mongrel2 architecture which could make the people switch?
>
> In my particular field, the answer is obvious, but what I am developing
> is a service, people will never know they use Mongrel2. So, I wonder and
> here are the things I would like to have at the moment:
>
> - a better Gmail? Think Mongrel2+Lamson coming together to people to do
> customer support or CRM or "friends and relatives" relation care.
> - a GitHub but distributed with OpenId and OAuth to fork/pull everywhere
> on 1000's of servers.
> - a better family/personal/company website? where you could have your
> code, your CRM, your pictures, your mailing list, etc.
>
> Just wondering, any idea? What are you missing? I do not have a pet
> project at the moment, all my projects are work related, so I wonder if
> some of the stuff I miss could be in line with what you miss. This could
> be the trigger to start something...
>
> Have a nice day,
> loïc
>
>
>
Re: [mongrel2] A killer app for Mongrel2
- From:
- David Kantowitz
- Date:
- 2012-01-01 @ 17:09
Folks, Apologies for the incoherent email ... hit forward instead of reply.
My buddy and I have an ongoing disagreement. He thinks PaaS is over-built
and done for excepting consolidations and bankruptcies. I think we're
barely seeing the first real "cloud platforms" arrive ... what we've seen
up to now is simply the ongoing historical oscillation between
compute-density vs. bandwidth.
Anyway, I think Ken is onto something with Mongrel2 being a good front-end
for a cloud platform.
Regards,
David K.
> On Sat, Dec 31, 2011 at 10:46 AM, Kenneth Keiter <ken@kenkeiter.com>wrote:
>
>> I think that a genuine 'killer-app' for Mongrel2 will be one which takes
>> advantage of what Mongrel2 specializes in. There are several situations in
>> which I would not use it because, frankly, there are better tools for the
>> job. But the best servers are opinionated in their design -- they tackle
>> one set of tasks incredibly well -- and Mongrel2 is no exception.
>>
>> Although *my* killer app is not open source, I take advantage of
>> Mongrel2's use of ZeroMQ (and frequently its ability to address many
>> clients with a single response) to flexibly scale client-facing
>> applications that I host. To do so, I've been developing my own PaaS
>> solution, which I call Tapestry. Here are some details:
>>
>> Tapestry's primary purpose is to allow simplified distribution and
>> automatic scaling of Ruby, Python, and Lua web applications/services on EC2
>> infrastructure. It uses a master/slave architecture in which a dedicated
>> EC2 machine with elastic IP is setup as a master, and runs Mongrel2. It
>> also acts as a hub with which slaves will communicate (via MessagePack RPC
>> spec over ZeroMQ). Slaves serve to support the modular web applications,
>> and are deployed automatically as new EC2 instances by the master. They
>> pull application code from a repo (using signed tags for verification), and
>> a bundled setup descriptor allows the application to reconfigure the slave
>> to support it. Here's the primary place where Mongrel2 comes in: the
>> Tapestry master reconfigures Mongrel2 to serve new applications
>> automatically, and the application instances running on separate slaves
>> connect back to the master's Mongrel2 instances via ZeroMQ over TCP to
>> scale the application horizontally. Multiple masters are allowed, and new
>> masters are advertised over mDNS. Voila!
>> Auto-horizontally/vertically-scaling (where appropriate), easily
>> distributed and updated web applications. Pretty simple.
>>
>> Anyway, my point is that there's a proper tool for the job in every
>> circumstance. If you're talking killer apps, I think the best choice is
>> going to be one that really illustrates how Mongrel2 gets out of the way
>> and makes serving/scaling incredible web applications into a clean and
>> simple process. We need an idea that illustrates:
>>
>> * ease of configuration (and reconfiguration on the fly);
>> * topological/architectural flexibility afforded by Mongrel2's use of
>> ZeroMQ;
>> * speed/ability to support vast numbers of clients;
>> * naught to up-and-running ease;
>> * ease of converting existing applications.
>>
>> Thoughts?
>>
>> Happy new year to everybody.
>>
>> Cheers,
>> Ken Keiter
>>
>> On Saturday, December 31, 2011 at 10:17 AM, Mike Laidlaw wrote:
>>
>> Not sure if this exactly qualifies as a killer app, but I've been
>> thinking about this problem: my sister-in-law and I were discussing the
>> hours of personal digital video that we have collected (kids-family
>> stuff) and what to do with it all.
>>
>> I was thinking it would be great if I could stick it all on a shared
>> server and then be able to access it with a youtube like interface from
>> any web-browser anywhere. The idea is we have total control over all of
>> the video, but we could use it just like youtube. Maybe there is
>> something like that out there already? Any thoughts?
>>
>> Happy New Year to all!
>>
>> Mike
>>
>>
>> On 10/12/2011 3:26 AM, Loic d'Anterroches wrote:
>>
>> Hello,
>>
>> no, I am not announcing a killer app for Mongrel2 which will make
>> everybody switch from Apache/Nginx to Mongrel2.
>>
>> I just wonder, what could be a killer app which could really take
>> benefit of the Mongrel2 architecture which could make the people switch?
>>
>> In my particular field, the answer is obvious, but what I am developing
>> is a service, people will never know they use Mongrel2. So, I wonder and
>> here are the things I would like to have at the moment:
>>
>> - a better Gmail? Think Mongrel2+Lamson coming together to people to do
>> customer support or CRM or "friends and relatives" relation care.
>> - a GitHub but distributed with OpenId and OAuth to fork/pull everywhere
>> on 1000's of servers.
>> - a better family/personal/company website? where you could have your
>> code, your CRM, your pictures, your mailing list, etc.
>>
>> Just wondering, any idea? What are you missing? I do not have a pet
>> project at the moment, all my projects are work related, so I wonder if
>> some of the stuff I miss could be in line with what you miss. This could
>> be the trigger to start something...
>>
>> Have a nice day,
>> loïc
>>
>>
>>
>
Re: [mongrel2] A killer app for Mongrel2
- From:
- Loic d'Anterroches
- Date:
- 2012-01-01 @ 18:05
Hello,
On 2012-01-01 18:09, David Kantowitz wrote:
> Folks, Apologies for the incoherent email ... hit forward instead of
> reply. My buddy and I have an ongoing disagreement. He thinks PaaS is
> over-built and done for excepting consolidations and bankruptcies. I
> think we're barely seeing the first real "cloud platforms" arrive ...
> what we've seen up to now is simply the ongoing historical oscillation
> between compute-density vs. bandwidth.
>
> Anyway, I think Ken is onto something with Mongrel2 being a good
> front-end for a cloud platform.
I am using it every day and I am more than pleased with it. I have
basically duplicated Heroku. Even reusing part of their "buildpacks".
This is working extremely well. At the backend level, ZMQ is basically
our lingua franca and it makes everything simpler.
I can only recommend M2 in such system, but now, I need to learn about mDNS.
Good evening,
loïc
> On Sat, Dec 31, 2011 at 10:46 AM, Kenneth Keiter <ken@kenkeiter.com
> <mailto:ken@kenkeiter.com>> wrote:
>
> I think that a genuine 'killer-app' for Mongrel2 will be one
> which takes advantage of what Mongrel2 specializes in. There are
> several situations in which I would not use it because, frankly,
> there are better tools for the job. But the best servers are
> opinionated in their design -- they tackle one set of tasks
> incredibly well -- and Mongrel2 is no exception.
>
> Although *my* killer app is not open source, I take advantage of
> Mongrel2's use of ZeroMQ (and frequently its ability to address
> many clients with a single response) to flexibly scale
> client-facing applications that I host. To do so, I've been
> developing my own PaaS solution, which I call Tapestry. Here are
> some details:
>
> Tapestry's primary purpose is to allow simplified distribution
> and automatic scaling of Ruby, Python, and Lua web
> applications/services on EC2 infrastructure. It uses a
> master/slave architecture in which a dedicated EC2 machine with
> elastic IP is setup as a master, and runs Mongrel2. It also acts
> as a hub with which slaves will communicate (via MessagePack RPC
> spec over ZeroMQ). Slaves serve to support the modular web
> applications, and are deployed automatically as new EC2
> instances by the master. They pull application code from a repo
> (using signed tags for verification), and a bundled setup
> descriptor allows the application to reconfigure the slave to
> support it. Here's the primary place where Mongrel2 comes in:
> the Tapestry master reconfigures Mongrel2 to serve new
> applications automatically, and the application instances
> running on separate slaves connect back to the master's Mongrel2
> instances via ZeroMQ over TCP to scale the application
> horizontally. Multiple masters are allowed, and new masters are
> advertised over mDNS. Voila!
> Auto-horizontally/vertically-scaling (where appropriate), easily
> distributed and updated web applications. Pretty simple.
>
> Anyway, my point is that there's a proper tool for the job in
> every circumstance. If you're talking killer apps, I think the
> best choice is going to be one that really illustrates how
> Mongrel2 gets out of the way and makes serving/scaling
> incredible web applications into a clean and simple process. We
> need an idea that illustrates:
>
> * ease of configuration (and reconfiguration on the fly);
> * topological/architectural flexibility afforded by Mongrel2's
> use of ZeroMQ;
> * speed/ability to support vast numbers of clients;
> * naught to up-and-running ease;
> * ease of converting existing applications.
>
> Thoughts?
>
> Happy new year to everybody.
>
> Cheers,
> Ken Keiter
>
> On Saturday, December 31, 2011 at 10:17 AM, Mike Laidlaw wrote:
>
>> Not sure if this exactly qualifies as a killer app, but I've been
>> thinking about this problem: my sister-in-law and I were
>> discussing the
>> hours of personal digital video that we have collected
>> (kids-family
>> stuff) and what to do with it all.
>>
>> I was thinking it would be great if I could stick it all on a
>> shared
>> server and then be able to access it with a youtube like
>> interface from
>> any web-browser anywhere. The idea is we have total control
>> over all of
>> the video, but we could use it just like youtube. Maybe there is
>> something like that out there already? Any thoughts?
>>
>> Happy New Year to all!
>>
>> Mike
>>
>>
>> On 10/12/2011 3:26 AM, Loic d'Anterroches wrote:
>>> Hello,
>>>
>>> no, I am not announcing a killer app for Mongrel2 which will make
>>> everybody switch from Apache/Nginx to Mongrel2.
>>>
>>> I just wonder, what could be a killer app which could really take
>>> benefit of the Mongrel2 architecture which could make the
>>> people switch?
>>>
>>> In my particular field, the answer is obvious, but what I am
>>> developing
>>> is a service, people will never know they use Mongrel2. So, I
>>> wonder and
>>> here are the things I would like to have at the moment:
>>>
>>> - a better Gmail? Think Mongrel2+Lamson coming together to
>>> people to do
>>> customer support or CRM or "friends and relatives" relation care.
>>> - a GitHub but distributed with OpenId and OAuth to fork/pull
>>> everywhere
>>> on 1000's of servers.
>>> - a better family/personal/company website? where you could
>>> have your
>>> code, your CRM, your pictures, your mailing list, etc.
>>>
>>> Just wondering, any idea? What are you missing? I do not have
>>> a pet
>>> project at the moment, all my projects are work related, so I
>>> wonder if
>>> some of the stuff I miss could be in line with what you miss.
>>> This could
>>> be the trigger to start something...
>>>
>>> Have a nice day,
>>> loïc
>
>
>
--
Dr Loïc d'Anterroches
Founder Céondo Ltd
w: www.ceondo.com | e: loic@ceondo.com
t: +44 (0)207 183 0016 | f: +44 (0)207 183 0124
Céondo Ltd
Dalton House
60 Windsor Avenue
London
SW19 2RR / United Kingdom