Sun Technologies for the Cloud

[Rao Dronamraju]

It is interesting to note that, from Sun’s technology building blocks for Clouds (http://www.sun.com/solutions/cloudcomputing/offerings.jsp),

Sun’s Grid Engine seems to be conspicuously missing.

 

What does this mean?....

 

Does Sun feel that the Clouds will be driven purely from a Hypervisor based resource management scheme (in addition to network and storage virtualization) as opposed to Grids which were OS based virtualization/resource management? ….

Does Sun feel Grid technologies of the past (by globus and other organizations) have been superseded or supplemented by the hypervisor technologies?....

 

Any comments from the Sun folks who have the insight into it?....

 

Thanks

Rao

 

[Kevin Apte]

Sun has announced that they are not accepting new customers for Grid Engine. Grid Engine is a high performance Computing solution not a cloud offering. 

 

Kevin

[Rao Dronamraju]

Yes, Sun did think about its Sun Grid Engine for its Cloud offering at one time but seems to have dropped it in favor of Q-Layer.

 

http://www.theregister.co.uk/2009/03/13/sun_cloud_supernap/

 

 

---

[Miha Ahronovitz]

Rao, very insightful observations. I will reply talking about Amazon
Web Services. Sun Grid Engine, is a resource manager, moving
resources to where they are most needed, and getting more resources
when demand increases

The public has a bizarre perception that by using AWS, we have
automatically a cloud. Many times we stated what is a cloud from a
user perspective is defined as follows:

1.A user will always have all resources s/he needs
2. A user will pay only for what it uses
3.The applications are delivered as an easy to use service
4.The users do not want to know what is going inside the cloud

Assume a single user opens an account,for AWS EC2 services. He places
his credit card and the instructions from AWS page are “simple”
Create an Amazon Machine Image (AMI) containing your applications,
libraries, data and associated configuration settings. … Upload the
AMI into Amazon S3. Amazon EC2 provides tools that make storing the
AMI simple... Use Amazon EC2 web service to configure security and
network access....

Choose which instance type(s) and operating system you want, then
start, terminate, and monitor as many instances of your AMI as needed,
using the web service APIs or the variety of management tools
provided....Determine whether you want to run in multiple locations,
utilize static IP endpoints, or attach persistent block storage to
your instances...

If I am not a developer, it gives me headache and I feel dizzy
already. Who will make the AMI's for me? Some products luckily have
some created, but what is I am not listed for the application I want?

This clearly violates the rule (4). The user must be really a
developer. Then assume s/he selected 3 nodes, but she realizes she
needed 10. The only way to adjust this is manually. That violates the
rule (1). Then, the user may need some way of balancing the load on
each virtual node. Thee is no built in tool in AWS to the job
equivalent to Sun Grid Engine (SGE)

A single user asking for AWS virtual resources is not operating a
cloud. He is creating a cloud. This implies AWS is not for users, it
is for developers.

Resource management tools like Sun Grid Engine are a big hope for AWS.
Or they should be. Yet AWS is not designed yet to work with resource
automation / management tools like SGE, which in theory could
automatically buy virtual resources from AWS, when demand increases.
This may mean huge business for AWS, but right now is not. SGE can
manage indiscriminately local physical resources and AWS resources,
and no one notices the difference. (sort of: in some applications the
latency from virtual nodes versus physical nodes can be a show
stopper).

But how the Amazon billing works? Can you imagine me placing my credit
card as a user of an SGE cluster, and the scheduler and Service Domain
Manager are accessing AWS without asking me? There is no way to place
a limit on my credit card and how do I know the scheduler decisions
meet my approval?

We can not know. Therefore no private users will entrust their credit
card and let a scheduler do their buying for them. This violates the
rule (2): A user will pay only for what it uses.

The only way to use a resource management manager, today is to have a
corporate credit card. Users pay nothing, but the IT that owns and
operates the cloud pays for it. There is no way, today, to have SGE
(or LSF or PBS, or GigaSpaces – Nati please let me know if I am
wrong) manage credit cards for individual users of their managed
cluster.

Yet, the way Amazon EC2 operates, is with individual credit cards.

Amazon EC2 is just a tool to build a cloud. It is not a cloud per se.
Soon, there will be stiff competence for virtual resource provider.
AWS are early adopters in operating a whole virtual resources business
in computing. That's commendable , but they a way to go still.

This way, my friends on this CC group, the business will be good. YOU
are the ones who will deliver the clouds, and make something usable
by end users. It is called Private Clouds, and the Private, Public and
Hybrid Clouds creation will mean a lot of business. SGE is synergistic
to cloud resource providers such as Amazon. They can prosper without
us.

2 cents,

Miha


On Mar 26, 6:02 pm, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>
wrote:

> It is interesting to note that, from Sun's technology building blocks for
> Clouds (http://www.sun.com/solutions/cloudcomputing/offerings.jsp),
>
> Sun's Grid Engine seems to be conspicuously missing.
>
> What does this mean?....
>
> Does Sun feel that the Clouds will be driven purely from a Hypervisor based
> resource management scheme (in addition to network and storage
> virtualization) as opposed to Grids which were OS based

> virtualization/resource management? ..

[Miha Ahronovitz]

Kevin, where did you get this erroneous  information from?  Please download Sun Grid Engine from here and  enjoy it for many years to come
http://www.sun.com/software/sge/get_it.jsp

Miha

---

From: Kevin Apte <kevi...@gmail.com>
To: cloud-c...@googlegroups.com
Sent: Thursday, March 26, 2009 6:49:08 PM

Subject: [ Cloud Computing ] Re: Sun Technologies for the Cloud

[David Levy]

Kevin

I think you'll find your language is inexact.

Rao was referring to the software product, known as Sun Grid Engine which is still available for download, for free and remains a  very popular management component of many of today's large scale clusters.

I am unclear as to what you refer about Sun not accepting new customers, but I assume it is the utility offering known by various names, including informally Sun Grid, which the cloud services announced at Community One earlier this month will supercede.

--

Dave

David Levy Chief Technologist Sun Microsystems Europe Global Sales & Services London, United Kingdom

Blog http://blogs.sun.com/DaveLevy Email David...@Sun.COM

Sun Proprietary & Confidential . This e-mail message is for the sole use of the intended recipient(s) and may contain confidential and priviliged information. Any unauthorised review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message.

[David Levy]

You have also become confused between the distributed resource manager, Sun Grid Engine and the time sharing service offering previously known as Sun Grid and later network.com.

network.com was optimised to service commercial HPC applications but didn't to my knowledge use Sun Grid Engine.

You pose Q-Layer and SGE as alternatives. I am not sure they need to be.

I have downloaded SGE this a.m. to see how it plays with OpenSolaris virtualisation technologies.

--

Dave

David Levy Chief Technologist Sun Microsystems Europe Global Sales & Services London

United Kingdom

Blog http://blogs.sun.com/DaveLevy Email David...@Sun.COM

[John Stanford]

Kevin,

To add to Dave's comment, Sun Grid Engine is not specifically listed as a cloud building block, however the product is still alive and well.  The building blocks listed are primarily aligned with building an IaaS offering.  As you can imaging many of the products in Sun's portfolio could be considered cloud building blocks if you are considering building a PaaS or SaaS offering (Grid Engine included).

Best Regards,

John 

[Rao Dronamraju]

Miha,

Thanks a lot for the detailed clarification.

Infact, I get the same impression with Microsoft's Azure, that it is
primarily designed for developers (atleast at this stage).

I do not understand the business rationale for building it for developers as
opposed to businesses or consumers unless they are saying let the Cloud
aware applications be built by the developers first and then they can be
enabled in the Cloud for use by the businesses....

But then in the initial stages of the cloud adoption (and may be even
beyond), it will be the existing applications that will be migrated to
Clouds because of all the investment both in monetary and technological
terms that has already gone into them.

Rao

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Miha Ahronovitz
Sent: Friday, March 27, 2009 1:38 AM
To: Cloud Computing
Subject: [ Cloud Computing ] Re: Sun Technologies for the Cloud

[William Louth (JINSPIRED.COM)]

"But then in the initial stages of the cloud adoption (and may be even beyond), it will be the existing applications that will be migrated to Clouds because of all the investment both in monetary and technological terms that has already gone into them."

And why exactly would companies be considering moving applications with "investment both in monetary and technological terms" to the cloud. Are we expecting the cloud itself to inject its magic into these applications without any risk and further costs (dev, test, ...)? There will have to be some change and not just in the software itself (products, processes, people) and this will have to be justified by cost reduction which can only achieved with greater change to the existing software (alignment to the cloud). One partial solution to this is to take an open architecture/platform used by existing applications and create an implementation that can inject such capabilities - i.e. a Java SE/EE server powered by the Cloud.

I think MS is looking down the road and without a hosting provider hat on and realizing that in the end it will all be about the software service (activity) itself and the closer it gets to observing & managing this the better chance it has of scaling its own cloud offering at significantly less cost than others who will be forced to push the cost direct or indirect back to the customer. Naturally some companies cannot wait around for new cloud services to be developed that run natively within some cloud runtime. They need money now while this evolves. MS and maybe some others have the luxury of focusing on the end game - creating an offering with less baggage and designed for a longer lifespan.

William

[Rao Dronamraju]

“And why exactly would companies be considering moving applications with "investment both in monetary and technological terms" to the cloud. Are we expecting the cloud itself to inject its magic into these applications without any risk and further costs (dev, test, ...)? “

 

Well, the very idea behind Cloud Computing is for businesses to “offload/outsource/CloudSource”  their data centers so that they do not have to run them themselves but the services will be provided by a CSP (Cloud Service Provider) at a much lower cost. Can a data center of a business be run today in a Cloud?...yes, it can be, from a point of view of availability of technology to provide same levels of services that they provide within their own data centers. So you can actually migrate a business today to a Cloud environment. The main problem is the companies are not comfortable with the Security, Privacy, Trust issues of their DATA. From an availability, scalability and performance perspective and even elasticity / autonomic/dynamic resource management perspective technologies are in place and available for CloudSourcing today’s businesses. Everything that can be done in an owned data center can also be done in an outsourced data center.

 

So the focus for making Clouds viable with business world should be in making the CXOs and CIOs of the businesses comfortable with technologies, products solutions that address Security, Privacy and Trust issues to bingin with not some application development on a platform that can be accessed over the web

 

---

[ademello]

Kevin, you are sorely mistaken on this point.

On Mar 26, 9:49 pm, Kevin Apte <kevina...@gmail.com> wrote:
> Sun has announced that they are not accepting new customers for Grid Engine.
> Grid Engine is a high performance Computing solution not a cloud offering.
>
> Kevin
>
> On Thu, Mar 26, 2009 at 9:02 PM, Rao Dronamraju <
>
>
>

> rao.dronamr...@sbcglobal.net> wrote:
> >  *It is interesting to note that, from Sun’s technology building blocks
> > for Clouds (http://www.sun.com/solutions/cloudcomputing/offerings.jsp),*
>
> > *Sun’s Grid Engine seems to be conspicuously missing.*
>
> > * *
>
> > *What does this mean?....*
>
> > * *
>
> > *Does Sun feel that the Clouds will be driven purely from a Hypervisor

> > based resource management scheme (in addition to network and storage
> > virtualization) as opposed to Grids which were OS based

> > virtualization/resource management? ….*
>
> > *Does Sun feel Grid technologies of the past (by globus and other

> > organizations) have been superseded or supplemented by the hypervisor

> > technologies?....*
>
> > * *
>
> > *Any comments from the Sun folks who have the insight into it?....*
>
> > * *
>
> > *Thanks*
>
> > *Rao*
>
> > * *

[ademello]

Consumers? I am still waiting for the business case for consumers to
use the cloud DIRECTLY.

Right now we see complex, HA apps like email and backups provided as a
service for consumers via a cloud. But would consumers build their own
email or backup service via a cloud? Makes no sense.

On Mar 27, 11:58 am, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>
wrote:

[ademello]

The data centers are a means to an end: the means through which
companies deploy their apps.

Clouds hide the complexity of the data center(s) and reduce
development complexity, and ensure a similar or better end user
experience.

What you are referring to Rao, is hosting. Rather than run a data
center for themselves, companies can outsource their data center to a
managed service provider. Exodus, way back in the day, had plans to
take over data centers for businesses everywhere. What they soon found
out was that managing heterogeneous applications and hardware was not
easy or cheap. Clouds solve that problem that standardizing the
infrastructure, and the platform(s).

Clouds are not here to make consumers deliver services to themselves.

Outsourced data center does not equal a cloud.

On Mar 27, 2:55 pm, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>
wrote:

> "And why exactly would companies be considering moving applications with
> "investment both in monetary and technological terms" to the cloud. Are we
> expecting the cloud itself to inject its magic into these applications
> without any risk and further costs (dev, test, ...)? "
>
> Well, the very idea behind Cloud Computing is for businesses to
> "offload/outsource/CloudSource"  their data centers so that they do not have
> to run them themselves but the services will be provided by a CSP (Cloud
> Service Provider) at a much lower cost. Can a data center of a business be
> run today in a Cloud?...yes, it can be, from a point of view of availability
> of technology to provide same levels of services that they provide within
> their own data centers. So you can actually migrate a business today to a
> Cloud environment. The main problem is the companies are not comfortable
> with the Security, Privacy, Trust issues of their DATA. From an
> availability, scalability and performance perspective and even elasticity /
> autonomic/dynamic resource management perspective technologies are in place
> and available for CloudSourcing today's businesses. Everything that can be
> done in an owned data center can also be done in an outsourced data center.
>
> So the focus for making Clouds viable with business world should be in
> making the CXOs and CIOs of the businesses comfortable with technologies,
> products solutions that address Security, Privacy and Trust issues to bingin
> with not some application development on a platform that can be accessed
> over the web
>
>   _____  
>

> libraries, data and associated configuration settings. .  Upload the

> (or LSF or  PBS, or GigaSpaces - Nati please let me know if I am

> wrong)  manage credit cards for individual users  of their managed
> cluster.
>
> Yet, the way Amazon EC2 operates, is with individual credit cards.
>
> Amazon EC2 is just a tool to build a cloud. It is not a cloud per se.
> Soon, there will be stiff competence for virtual resource provider.
> AWS are early adopters in operating a whole virtual resources business
> in computing. That's commendable , but they a way to go still.
>
> This way, my friends on this CC group, the business will be good. YOU
> are the ones who will deliver  the clouds, and make something usable
> by end users. It is called Private Clouds, and the Private, Public and
> Hybrid Clouds creation will mean a lot of business. SGE is synergistic
> to cloud resource providers such as Amazon. They can prosper without
> us.
>
> 2 cents,
>
> Miha
>

> On Mar 26, 6:02 pm, "Rao Dronamraju"  <mailto:rao.dronamr...@sbcglobal.net>

[Rao Dronamraju]

Why don’t you use Google Search?....Is Google Search a Cloud
Application?...assuming Google's data centers dynamic and elastic in cloud
resource management.

Tomorrow if Microsoft hosts its Office Apps on Azure and Intuit hosts its
widely used apps, don't you think these are consumer (as in consumer markets
as opposed to SME) SaaS apps in a Cloud?....

"But would consumers build their own email or backup service via a cloud?
Makes no sense."

Where did you get the impression that I said consumers are going to build
their own applications on the cloud, I was saying consumers using Cloud
applications or clouds providing SaaS for general consumers.

[Jeanne Morain]

Nicely put. The Cloud is still very Ambigious - is it a Web 2.0 Application, Private Cloud - one central unit in a company hosting all applications to be used by ALL business units, Software As A Service, Desktops as a Service or somewhere in between?

The CXOs need to be comfortable with the GRC aspects (Governance Risk and Compliance) as well as the economies of scale. For smaller companies external hosted clouds or solutions like Salesforce.com make more sense but larger companies have more to track and therefore more risk.

It is still too early to tell where this will end up but many vendors other then Microsoft are focused on the end game - a hybrid solution that is about the fundamental pieces that mean the most to CXOs - User Productivity and GRC... both connected and disconnected....

[Rao Dronamraju]

Adamello,

So what is Cloud Computing?...As we all know, the term originated from the
internet cloud that we all are used to for the last 10+ years.

The computing environment/eco systems will be like a cloud for the user?..
Which user?...Consumer?...business?....or both?....

I think it is both.

So for a general consumer like you and me, Clouds have been happening
already. It is for the businesses that Cloud Computing is a major shift. It
started with consumer markets and progressing to SME markets.

If a business uses its IT services just like a consumer uses the internet
today where all the complexities of data center are hidden from him/her, I
think that is Cloud for the businesses. So basically a business instead of
owning and building its data center will get those services from a provider.
How can a provider provide such services to a business?....by more or less
operating the same data center in his/her premise. So is this
outsourcing?...or cloud?...IMO, it does not matter as long as the user
whether a consumer or a business has the cloud like experience of the
internet.

"Clouds hide the complexity of the data center(s) and reduce
development complexity, and ensure a similar or better end user
experience."

Depends on which user you are talking about, again consumer or business?...

If it is a consumer....

Today's data centers also hide the complexity of applications to its
users/consumers.
How about online banking?....Do you see any of the complexity involved with
supporting data center infrastructure, middleware, applications etc when you
are using online banking?.. If the online banking application is moved to a
cloud, would the user experience be any different?...

In case of businesses as long as the outsourcing provider provides the
following as Miha mentioned in an earlier email, and I agree with him, what
difference does it make if you call him and outsourcer/managed
services/cloud service provider?..

1.A user will always have all resources s/he needs
2. A user will pay only for what it uses
3.The applications are delivered as an easy to use service
4.The users do not want to know what is going inside the cloud

"Clouds are not here to make consumers deliver services to themselves."

I am not sure at all where you got this from and what exactly you mean by
this?...

[Bert]

>>>Many times we stated what is a cloud from a user perspective is defined as follows:

Who is "we" in this context?

I'm curious because you seem to be limiting your discussion of cloud
computing to a narrow consumer scope by insisting users do not want to
know what is going on inside the cloud. I can assure you that at
3tera, our users very much insist on knowing because they run their
businesses in the cloud.

Bert

[ademello]

Where did I get that idea? Here;

"Infact, I get the same impression with Microsoft's Azure, that it is
primarily designed for developers (atleast at this stage).
I do not understand the business rationale for building it for
developers as

opposed to businesses or consumers..."

Consumer don't need "clouds." Consumers don't care. Your Google
example is a good one. Nobody cares if Google search is run on a
cloud, a mainframe, or Larry Page's laptop. The end user experience is
seamless.

So, as to your question re: Inuit / MS Office, consumers don't care if
its run on a cloud. They may care if its software as a service, since
they don't have to buy shrinkwrap or install / maintain an app, and
have to trust the SaaS provider with their data.


On Mar 27, 9:51 pm, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>

[ademello]

Rao,

You keep talking about consumers having used "the cloud" for years.
From your first sentence, you seem to imply that the term "cloud" came
from the Internet cloud - hence, consumers have been using "the cloud"
for years.

We are not talking about "the cloud" here, but cloud computing. I
define cloud computing as elastic, pay-as-you-go computing resources,
with a value-added stack on top (providing HA, linear scalability,
automatic backup, etc.) , all accessed by a simple API. In theory,
apps running on the same cloud should be able to share data
synergistically.

From your very own example about online banking, consumers don't care
about clouds. They may care about services delivered via a cloud
(email, backup, productivity apps) but really, they don't care about
the clouds behind the services, much as when I visit a retail bank I
don't concern myself with the IT that keeps that bank running.

Cloud computing is not about consumers, its about IT and business
adapting to a world where they no longer need to worry about the core
services that form the backbone of IT today. On this point we seem to
agree.

From my soapbox, one o the biggest wastes of money in any fairly sized
company, is their IT staff. The entire IT business has grown on the
back of an extremely inefficient and unnecessarily complex "value
chain" that has existed almost entirely to benefit those providing the
IT services. The fact that this industry has grown so incredibly
stands testament to the benefits we have seen from automation via IT
as a society.

Fortunately, (or unfortunately if you are on the wrong side of
Darwin's evolution chart this time around), this is all going to come
to a screeching end very soon. Not only do I feel that its being
accelerated by the global economic situation, its being accelerated by
market competition itself.

A new business starting today does not have to deal with the baggage
of old infrastructure. From my own world, we saw this with Reliance in
India, and China Mobile. For them to grow at the rates they did, they
had to innovate from within - the market could not provide the scale
of the services that they needed in certain cases.

The efficiency at which they run their business is breathtaking if you
compare it to a Verizon, or AT&T. Or even Vodafone.

This very same efficiency revolution is going to hit the business
world big time in the next 3-5 years. Part of it will come from
necessity to cut costs, part of it will come from the need to compete
with much more lithe, agile companies.

The major challenge, as Jeanne hinted at, is the CxOs are going to
fight desperately against it in many cases, citing GRC issues, or
other canards. Salesforce managed to break through this barrier, and I
think has paved the way for a broader cloud computing industry.

Lastly, I want to mention that I'm not a "general consumer" as you
said I am. I have a vested interest in this, as I own and run a
company developing the building blocks for business and governments to
operate their own private clouds,

Best,

Aaron

On Mar 27, 11:14 pm, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>

> ...
>
> read more »

[Rao Dronamraju]

"I define cloud computing as elastic, pay-as-you-go computing resources,
with a value-added stack on top (providing HA, linear scalability,
automatic backup, etc.) , all accessed by a simple API. In theory,
apps running on the same cloud should be able to share data
synergistically."

Everything you have mentioned in the definition of cloud, can be done in
today's data centers. A data center can be elastic today, infact quite a few
companies have been working on Grid/Utility/On Demand Data Centers for
atleast 10 years, they are also leasing servers on a pay-as-you-go model,
many of them especially in the enterprise world have HA, linear scalability
automatic backup etc. The software that provides the above services all of
them have API's that you can use. Infact in today's data centers all apps
can share data based on the security configurations. So what is it in your
definition of clouds that is different from today's data centers?...

I think the big difference is in the ownership of the data center. If you do
not like the word data center let me use the word cloud center.Just like a
general consumer does not care what is happening in the internet cloud that
is providing the services to him/her, a business/corporation also in a very
ideal case does not want to know what and how things are happening in the
Clouds and they also would like their IT services provided to them just like
non-business consumers are being provided with services over the
web/internet today. So basically an IT center of a business will get all its
services delivered to it (ERP services, security services, billing services,
business intelligence services etc). Much like payroll services are
completely delivered today to many businesses from a payroll services cloud.

"From my soapbox, one o the biggest wastes of money in any fairly sized
company, is their IT staff. The entire IT business has grown on the
back of an extremely inefficient and unnecessarily complex "value
chain" that has existed almost entirely to benefit those providing the
IT services. The fact that this industry has grown so incredibly
stands testament to the benefits we have seen from automation via IT
as a society."

So what is the solution?....you are going to run your operations that
provide the IT services in a cloud, is it not?...So the next question is do
you own the cloud (private) or do you run it in a public cloud?...If you own
the cloud you still have the CapEx and OpEx. If you do not own the cloud,
then you have solved your above problem, which is not owning the cloud. Now
, is this outsourcing?.....

"The major challenge, as Jeanne hinted at, is the CxOs are going to
fight desperately against it in many cases, citing GRC issues, or
other canards. Salesforce managed to break through this barrier, and I
think has paved the way for a broader cloud computing industry."

I think it is going to be a slow incremental stage by stage adoption of
cloud might happen. It is not easy to over the come the FUD when you as a
CXO is responsible for multi-million dollar IT data center or even
multi-million/billion dollar business is dependent IT. Here is a very
interesting real-life case that explains the FUD of the CXOs.

http://www.cio.com/article/101505/When_Bad_Things_Happen_to_Good_Projects

If SAP a well known technology can cause so many problems, Cloud Computing
which is far more sweeping and disruptive can be justifiably viewed with
much apprehension from a CXO perspective. Initially non-critical IT services
will be cloud sourced and slowly as CXOs gain confidence more and more
critical services will be obtained from a cloud.

[Miha Ahronovitz]

Rao, Aaron,

> 1. A user will always have all resources s/he needs

> 2. A user will pay only for what it uses

> 3. The applications are delivered as an easy to use service
> 4. The users do not want to know what is going inside the cloud

The corollary of the above is that Data Center, outsourced or not, is
transforming - for its owners - from a cost center, into a profit
center, once it becomes a cloud.

This is major shift that started with Amazon Web Services sending
an INVOICE. This natural consequence of having resources available
outside the organization - like electricity,virtual nodes, storage
and so on - is that the IT Director receives an invoice, and s/ he
does not know who should pay for it.

Why? Because the IT department, does not invoice traditionally the
employees and the company affiliated users. The major resource
management software tools today simply don't have a billing module.
The only possibility is the IT department pays for this invoice from
AWS and adds it up to budget they should request next year.

What will happen short term is we will add internal billing modules
inside the corporate ITs , able to interface to external billing
systems (like AWS) and -as an option - assign charges to the internal
project that generated the need for this external resources.

The idea of billing internal users is not as heretical as it sounds.
The users will not pay in cash the invoice. But the IT will know for
sure who consumes what and will be able to report back how much it
cost the to offer services to the users.

If one event will declare victory in the cloud computing revolution,
is the introduction of an internal billing in every IT department,
which, from that moment onwards will operate a cloud, not a data
center.

This is a huge qualitative jump: the cloud ecosystem will evolve with
private, hybrid, public clouds using many wholesale resource providers
(like AWS) to offer lower and lower cost services with increasing
quality of service through competition and benefit ultimately the
users of the clouds worldwide.

This is not a vision, but a real, achievable state. Here are some idea
for business start-up

- Internal Billing System Modules
- Professional services to restructure organizations budgets for IT
billing
- Remote billing services for internal IT
- Resource Optimization in a cloud eco-system
- job scheduling using pricing as one more criteria
- Porting applications to clouds
- become billing aware
- create AMI (Amazon Machine Images) and AMI-like for others
- ... many more

Cheers,

Miha

On Mar 28, 9:19 am, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>
wrote:

> http://www.cio.com/article/101505/When_Bad_Things_Happen_to_Good_Proj...

> ...
>
> read more »

[Miha Ahronovitz]

Bert, I should have added, "the users from the bucket you use as a target". I have seen the 3tera demo, which is clearly addressed to a bucket called "developers".

You state: "all you have to do is to drop the code in here". How helplessly simple, but not for Jo Shmo  the user.

What you call a cloud is only for a bucket containing a niche of relatively elitist developers placing apps to be delivered as a service to what I call a cloud.  The bucket I refer to  is full of Jo Shmo's interested to use the application as a service. That's all.

Your audience for simplicity is "developers". This does not rest merit to your start up, but you are not the end users cloud, you are part of the cloud ecosystem.

Miha

---

From: Bert <bar...@3tera.com>
To: Cloud Computing <cloud-c...@googlegroups.com>
Sent: Saturday, March 28, 2009 3:38:02 AM

Subject: [ Cloud Computing ] Re: Sun Technologies for the Cloud

[Greg Pfister]

Not really replying to any single person, just noting that this is

Yet Another Thread Chasing Its Tail Over the Definition of cloud
computing

(YATCITOD)

Greg Pfister
http://perilsofparallel.blogspot.com/

[ademello]

Miha,

In 1999 I built an internal billing system for a unit of Nortel. The
idea was to have an internal chargeback model whereby the company's
centralized IT would become a profit center, or at least run at break-
even.

The company had many visionary VPs, and the one who sponsored our
project was all into utility computing. The software we designed
worked fine but changing the processes at a large company is a HUGE
HUGE challenge. Huge.

Getting invoices from IT or for IT isn't going to be warmly embraced.

I do generally believe that the CIO should run IT as a service unit to
the company and that it should break even. Unprofitable and useless
projects should somehow be compensated for. Lastly, the people who
work in IT should deign to make themselves useful and profitable for
the company.

> ...
>
> read more »

[ademello]

Rao,

So the big revolution with AWS is that Amazon owns the data center?

You continue to mix SaaS and "clouds" or "cloud computing." This is
extremely confusing and does a disservice to the point you are trying
to make. ADP provides a solution via SaaS that allows you to
streamline your payroll. It is not a "payroll cloud." Maybe the
situation is different in the USA or where you are, but one cannot go
develop and deploy applications on ADP's systems, therefore it is not
a "payroll services cloud."

If your first paragraph, you make a series of statements that to me,
have completely no basis in reality. Like:

"The software that provides the above services all of them have API's
that you can use. Infact in today's data centers all apps can share
data based on the security configurations."

Where? In today's data centers I have no idea of the data model or
even the services running on other machines. Even in very large
companies with hugely sophisticated IT staff and excellent
outsourcing, I have yet to see anyone develop anything approximating
an API to access the data contained within the silos in the company.
Forget about sharing resources!

I mentioned in another post a long time ago that Exodus tried this
back in the Internet days. I worked very closely with them on a number
of big deployments. It was an extremely ambitious plan to host and
manage apps for customers in giant scale, and it was a failure. Partly
because the business model wasn't right, partly because the cost was
too high, and partly because the technology was immature. In today's
world, all of this was possible.

More recently, LoudCloud tried this. They essentially rebuilt the
entire Internet stack and added provisioning, etc. They were sold to
HP and now live in HP's managed services dept. What they did was
revolutionary for the time, and certainly not commonplace. Much of
what has happened since has been inspired by their hard work and deep
thinking.

Lastly, if you look at what Rackspace is doing with Mosso, that's the
evolution of the managed data center space. Rackable to Mosso. Its a
new service from their customer's point of view, as it allows them to
stop worrying about boxes. Rackable = boxes. Mosso = cloud.

Auto-scaling and HA are some of he key value-added features of cloud
"hosting" companies such as GoGrid and Mosso. You seem to think that
this is old hat and the real revolution is who owns the boxes, but I
disagree

Cheers,

Aaron

On Mar 28, 12:19 pm, "Rao Dronamraju" <rao.dronamr...@sbcglobal.net>
wrote:

> http://www.cio.com/article/101505/When_Bad_Things_Happen_to_Good_Proj...

> ...
>
> read more »

[Rao Dronamraju]

I agree.

Infact, here is a way we may get a closure (hopefully) on the definition of
Cloud or may be more controversy?..

Representatives from each of the big companies, who have all the resources
and staff to do dedicated full time research on Cloud Computing, who must
have already spent enough time on it and have arrived at their own
definition/conclusions about cloud, might want to help out.

So each of the representatives from HP, IBM, MS, Sun, Dell, Google, Yahoo,
Oracle, Intel (and any other who would like to join) if they are on this
forum may want to post their company's definition of cloud.

Interested in seeing if there is any consensus on this subject.

It will be great if they can post links to their direction/strategy and
roadmap if one exists.

Thanks

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Greg Pfister
Sent: Saturday, March 28, 2009 4:10 PM
To: Cloud Computing
Subject: [ Cloud Computing ] Re: Sun Technologies for the Cloud

[Matthew Zito]

Let's be very clear about LoudCloud.  LoudCloud was a "IT as a utility" company, where they saw the ability to leverage automation as a key differentiator vs. traditional managed hosting providers that did everything by hand.  LoudCloud built a software stack, and sold their services to web companies.

However, post-2000-downturn, their web business dried up with many of the dot-coms.  While a publicly traded company, they were losing significant amounts of money, and had to lay many people off.  The best feedback they were getting from the organizations that were still spending money, such as finance, government, etc., was that while they loved the idea of leveraging automation for improved IT efficiency, there was no way they were going to put their sensitive data in a third-party's datacenter.

So, LoudCloud worked out a very complicated arrangement with EDS.  EDS invested in LoudCloud, and also purchased an enterprise site license for LoudCloud's software.  In exchange, LoudCloud gave EDS all of their facilities, changed their name to Opsware, and entered the direct software market.

Now, as Opsware, they went back and started selling their software to the same organizations who were interested in the concept, but wanted an on-premise solution.  What's interesting is that while Opsware's software products could have drifted towards "on-demand IT", it really became focused around configuration consistency and standards compliance. 

That is, while LoudCloud had the revolutionary concept of "IT as an on-demand utility", what ended up making them successful was the fairly prosaic idea of "hey, when we build a server, it ought to be identical" and "I'd like to see everyplace where I've got servers missing security patches". 

In the end, Opsware was acquired by HP for $1.6b, approximately 11x revenue (I'm remembering that off the top of my head, could be wrong).  HP was one of a number of serious suitors for the technology.  Their biggest competitor, Bladelogic, was acquired by BMC for ~$800m.

A year later, HP acquired EDS.  HP runs EDS and the Opsware folks as totally separate divisions, though EDS still uses Opsware for certain types of work.

I guess the overall point I'm making, above and beyond clarifying your statement below, is that LoudCloud was revolutionary but unsuccessful.  It was the very boring "meat and potatoes" work of provisioning servers, compliance, reporting, etc. that made them a 1.6bn success.  Along the same lines, while I think a lot of the cloud concepts are revolutionary, I wouldn't be surprised if the "meat and potatoes" work of running shared compute environments is where the money will be made.

Thanks,
Matt

[ademello]

Matt - I couldn't agree more. I wasn't going to be the one who said
that LoudCloud wasn't successful... glad you said it for me. ;-)

Marc Andreesen made the point very well on his Charlie Rose interview.

Most of the money made in the next 18-24 months is definitely going to
be in the "meat and potatoes" backbone of cloud computing, and for
myself and other vendors in this space, that largely means selling a
cloud vision and strategy with some basic building blocks available
today.

> ...
>
> read more »

[Ray Nugent]

Matt, I couldn't agree more. At the end of this, it will most likely be about data center automation. Now, some of that automation may be to provision and balance work into a cloud and some of it may be to make an enterprise data center operate somewhat like a cloud, but the biggest cost in running a data center is still labor so that's the problem that will get solved.

Ray

---

From: Matthew Zito <mz...@gridapp.com>
To: cloud-c...@googlegroups.com; Cloud Computing <cloud-c...@googlegroups.com>
Sent: Saturday, March 28, 2009 5:36:00 PM

[Frank D. Greco]

At 02:33 PM 3/28/2009, Miha Ahronovitz wrote:
>The idea of billing internal users is not as heretical as it sounds.
>The users will not pay in cash the invoice. But the IT will know for
>sure who consumes what and will be able to report back how much it
>cost the to offer services to the users.

From the "There are More Albino Tigers than True IT innovations" Dept:

Billing internal users for usage was done in the late 70's and 80's.

Recall that Unix has had an accounting subsystem practically
since its creation. IBM mainframes had fairly sophisticated resource
accounting as well which supported several cloud^H^H^H^H^Htimesharing
vendors of that era.

The military used Unix accounting to determine how much computing
resources went into each project. Unix users had to use different
logins when working on different projects so their usage could be
tracked for budgetary and capacity planning purposes.

>This is a huge qualitative jump: the cloud ecosystem will evolve with
>private, hybrid, public clouds using many wholesale resource providers
>(like AWS) to offer lower and lower cost services with increasing
>quality of service through competition and benefit ultimately the
>users of the clouds worldwide.

Its nice to have a vision, but unfortunately reality often gets
in the way. True utility computing is *years* away imho. We will
eventually get there, but there's going to be a lot of other stuff
happening before that milestone. Its way early in the cloud era.
Look at how everyone in this forum defines the term differently.
And we're supposed to be the experts... But I guess chaos is not
unexpected in the early parts of a wave.

There's probably going to be a lot of clouds... different kinds.
Not just basic iron and bag services. The cloud stack is still
in its infancy. Utility computing will be there, but imho, there
are other areas that will crest earlier.

Btw, some of us are assuming the utility computing follows the
electricity model. This may not be true. Besides, the current
power grid in the US is abysmal; I'm not sure its a good
model to copy.
The analog-to-digital conversion sometimes can be lossy... :)

Frank G.

[Raja Srinivasan]

I've heard Cloud Computing compared to Electric Utilities.But it not there yet. Electric Utility has the following 3 major features:
*  Available for anybody willing to pay for services anywhere in the country (at least in the US).
*  Common payment method based on  Kilowatt hour -- usage model. The charge is the same no matter what the source of the electric power (Hydroelectric, nuclear, gas, oil, wind, solar)
*   A common interface for all appliances -- voltage, amps etc....

For Cloud Computing to be like a utility it must have these 3 major features.
*  Availability: One could argue that we have this today. Anyone with access to the web is leveraging Cloud Computing.
*  Billing. We need to come up with a common unit for billing that comprises both CPU usage, storage usage, as well as bandwidth consumed -- sort of Hertz-Byte-Hour. This has to be averaged out to consider all usage patterns Transaction Processing, OLAP etc. We are no where close to doing this.
*  Standard Interfaces: At this time the only applications I see on Cloud Computing are those related to personal productivity. This is a start, but we also need to develop application templates that are easy to deploy in any cloud vendor of my choice -- be it Amazon or Microsoft or Google or anyone....

Still have a long way to go in Cloud Computing.....

Thanks & Regards
Raja Srinivasan

[Tim M. Crawford]

Is labor really the largest cost in a data center? It depends on what you consider the 'data center'. By most measures, the largest operational cost for a data center is either the capital costs or energy costs.

-t

 

_________________________________________
Tim M. Crawford
+1.650.804.1300

timmcr...@yahoo.com

http://timcrawford.wordpress.com/

---

From: Ray Nugent <rnu...@yahoo.com>
To: cloud-c...@googlegroups.com
Sent: Saturday, March 28, 2009 6:27:20 PM

[Tarry Singh]

I fully agree that it is a long way to go before we can do realtime billing.metering and also provide mean rates (similar to how a typical hybrid car , Prius works - speed upto 50 km per hour - electricity, above 50 km/hr Gasoline - benzine? Diesel? other?), something that can be charged typically to consumers. Avarage consumer electricity rates vary across the globe can be used as

Availability - Honestly we aren;t there yet but there are many reasons behind it  and they are not all infra reasons. Adaptive computing is one thing but adaptive consumption is another. The latter is very hard to encompass as human masses will define when and what must be available. As social networks begin to evolve and start to congregate, it will be hard to plan way ahead in advance. And obviously the agile and realtime cloud has to still be delivered. current failures of salesforce, goog, azure are all evidence that we have a long way to go.

Metering: I'm writing a paper on metering and want to propose that it is time to use the expertise and intelligence of already existing utility industries such as water, electricity grids and take that as a standard to measure and bill the consumers. and yes again you're quite right, we are nowhere close to doing it in a standard and acceptable fashion. For the same reason there is a body being erected in EU to ask data centers to comply with some (now voluntary but later mandatory) CoC (Code of Conduct) where I will be proposing a SIGnaturing methodology for measuring the utility index/compliance standard. And that is what I've mentioned in previous thread as a RCC (Regulated Cloud Computing), or else I sincerely fear that several "toxic products" (such as 20 years insurance , AAA status, etc)  will emerge and eventually strike same kinds of deals that has cost the finance industry its credibility and very own existence.

The third will be the hardest I guess, but you've mentioned some very crucial points.

Tarry

--
Kind Regards,

Tarry Singh
______________________________________________________________
Founder, Avastu Blog: Research-Analysis-Ideation
"Do something with your ideas!"
Business Cell: +31630617633
Private Cell: +31629159400
LinkedIn: http://www.linkedin.com/in/tarrysingh
Blogs: http://www.ideationcloud.com

[William Louth (JINSPIRED.COM)]

Cloud computing is not a pure utility model. It is more like a hybrid
utility-service model.

I am not sure why we think a single common unit is required as this
assumes that all the activity of a cloud application or service will
exist within one cloud platform and use only use (primarily) one
computing resource. I expect that we will see cloud applications (user
facing services) running on one cloud platform but utilizing various
metered services both internal and external to the cloud platform. Each
of these metered services will naturally bill at a unit that is more
appropriate to activity characteristics of the delivered service and not
at the cost (and its resource usage) of the implementation of the service.

I recently published some slides from a recent cloudcamp talk I gave to
demonstrate how this is somewhat differently that a utility model.

Cloud Computing - A Tale of Two Machines
http://williamlouth.wordpress.com/2009/03/23/cloud-computing-a-tale-of-two-machines/

Some rather loose definitions I use:

Cloud Application - a user facing cloud service managed and metered by
an application provider and deployed to a single cloud platform.

Cloud Service - software offered up via a programmatic interface for the
purpose of constructing cloud applications. The service is managed and
metered by the service provider and exists in a different cloud platform(s).

Cloud Software - software that has been designed to be deployed to one
or more cloud platforms. The software is deployed and executes solely
within the platform of the cloud application. The software is managed
and possibly metered by the provider of the cloud application.

William

[Randall Minter]

We can find the answer in this paper from January 2009:

The Cost of a Cloud: Research Problems in Data Center Networks
http://ccr.sigcomm.org/online/?q=node/436

Quote:

Amortized Cost Component Sub-Components

∼45% Servers CPU, memory, storage system
∼25% Infrastructure Power distribution and cooling
∼15% Power draw Electrical utility costs
∼15% Network Links, transit, equipment
Table 1: Guide to where costs go in the data center.

It is natural to ask why existing solutions for the enterprise
data center do not work for cloud service data centers.

First and foremost, the leading cost in the enterprise is opera-
tional staff. In the data center, such costs are so small (under 5% due
to automation), that we safely omit them from Table 1. In a well-run
enterprise, a typical ratio of IT staff members to servers is 1:100.
Automation is partial [25], and human error is the cause of a large
fraction of performance impacting problems [21]. In cloud service
data centers, automation is a mandatory requirement of scale, and
it is accordingly a foundational principle of design [20]. In a well
run data center, a typical ratio of staff members to servers is 1:1000.
Automated, recovery-oriented computing techniques cope success-
fully with the vast majority of problems that arise [20, 12].

:Quote

- Randall

Founder & CTO
Qrimp, Inc.

http://www.qrimp.com

[Miha Ahronovitz]

Aaron, good to know.

"I CAN is 100 time more important than IQ", someone said.

In 1999, billing was a management idea, not dictated by an ecosystem like the cloud. There were no virtual resources to buy outside the organization, like in 2009

The psychological reaction inside the  Enterprise is important. I would start by just making  the $ charges and publish them on an internal web page, accessible to everyone inside.

Your experience in building an internal billing system could make you prosperous, if you start up a company today and execute well.

Miha

---

From: ademello <adem...@gmail.com>
To: Cloud Computing <cloud-c...@googlegroups.com>

Sent: Saturday, March 28, 2009 4:07:27 PM
Subject: [ Cloud Computing ] Re: Internal Billing for Cloud Computing Re: Sun Technologies for the Cloud

[Miha Ahronovitz]

Frank,

        True utility computing is *years* away imho.  We will
        eventually get there, but there's going to be a lot of other stuff
        happening before that milestone.

I am not sure we can postpone this, while AWS is charging daily hundreds, perhaps thousands of credit cards. It is not something optional to have a billing system. It is a necessity built in the definition of the cloud. In any organization using AWS, we need to handle these charges, and find out who generated them and decide who is responsible to pay them,

Miha

---

From: Frank D. Greco <fgr...@javasig.com>
To: Cloud Computing <cloud-c...@googlegroups.com>
Sent: Saturday, March 28, 2009 8:03:14 PM
Subject: [ Cloud Computing ] Re: Internal Billing for Cloud Computing Re: Sun Technologies for the Cloud

[Miha Ahronovitz]

William, I saw your presentation. It is dated March 23, and it shows the idea of billing, is not in my mind, or your mind, it is on our minds.

Miha

Post Scriptum:

C.J. Jung calls this "bio-culturally transmitted content of humanity’s collective unconscious"

Stanford University is only a few  miles away from my office . There is this an inscription on the Wall of the West Transept of Stanford's Memorial Church

"Thoughts and words travel just as God's life travels, They do not travel like an individual, but you breathe your spiritual life into the atmosphere as you do your breath, and some one else breathes it in. Those not present still receive it, for it permeates space, and all live in it and receive from it according to their unfoldment."

---

From: William Louth (JINSPIRED.COM) <willia...@jinspired.com>
To: cloud-c...@googlegroups.com
Sent: Sunday, March 29, 2009 3:27:07 AM
Subject: [ Cloud Computing ] Re: Internal Billing for Cloud Computing Re: Sun Technologies for the Cloud

[Jim Starkey]

Ray Nugent wrote:
> Matt, I couldn't agree more. At the end of this, it will most likely
> be about data center automation. Now, some of that automation may be
> to provision and balance work into a cloud and some of it may be to
> make an enterprise data center operate somewhat like a cloud, but the
> biggest cost in running a data center is still labor so that's the
> problem that will get solved.
>

Labor? That's an unusual take. Could you explain?

In the past, there has been a great deal of confusion over
administration costs. For enterprise level data centers, administration
costs include the cost of administering applications and the servers in
which they run. For very large data centers (aka "the cloud")
administration is limited to running physical servers hosting virtual
machines; the cost of administering the virtual servers and applications
is performed elsewhere, and isn't counted.

So where is in the labor? Is it running the data center, per se, or
administering applications and systems?

[Rao Dronamraju]

Miha,

 

I like your quote “I CAN is 100 time more important than IQ”

 

But I CAN + IQ could be 100++ times more important…

 

Regards,

Rao

 

 

---

[Greg Pfister]

I agree, and it's what immediately stood out to me in the paper
Randall's post pointed to.

Physical server hosting costs are grossly reduced by strict
homogeneity -- everything is exactly the same hardware base and OS
release (or a few of them). That's true of new installs, and
paradigmatically true of Google. I personally know of large a large
HPC installation with 1000s of nodes that runs with a total of 12
staff -- including people helping with application tuning. They do
this by being absolutely obessive about nodes being identical.

That's the space new clouds occupy (if they're smart), and by tossing
application and subsystem management over their shoulders at clients,
they can have really low management costs.

Long-run corporate data centers, on the other hand, regularly have a
conglomeration of random gear from consolidating previously
independent departments and acquisitions. There are often many
different hardware platforms, and even on the same hardware there
regularly are different OSs and other infrastructure. Too often they
run business-critical applications that will fall over on anything
else. All the different areas of expertise needed wind up requiring a
large number of people to cover it all.

So we have to be careful what we're comparing here, or we'll be mising
apples and 12-year-old single-malt scotch.

Greg Pfister
http://perilsofparallel.blogspot.com/

[Rao Dronamraju]

I started by eating my own cookie.....

So I went to IBM site and found the following information.....

http://www.ibm.com/ibm/cloud/

The white paper "Seeding the Clouds: Key Infrastructure Elements for Cloud
Computing" is decent.

Interestingly, they do not think Clouds are a logical extension of Grids
(atleast that is the impression I get).

One fundamental difference I am seeing between Grids and Clouds is in the
underlying resource architecture.

Grids are more tightly coupled structured clusters of clusters.

Clouds are fabric based architecture where loose coupling and unstructured
resources are the fundamental building blocks.

I will post info from each one of the companies I mentioned as and when I
get an opportunity...in addition to what might be posted by our friends.

Regards,
Rao

-----Original Message-----
From: cloud-c...@googlegroups.com
[mailto:cloud-c...@googlegroups.com] On Behalf Of Rao Dronamraju
Sent: Saturday, March 28, 2009 7:06 PM
To: cloud-c...@googlegroups.com

[Ray Nugent]

Yes.

---

From: Jim Starkey <jsta...@NimbusDB.com>
To: cloud-c...@googlegroups.com
Sent: Sunday, March 29, 2009 9:51:44 AM

Subject: [ Cloud Computing ] Re: Sun Technologies for the Cloud

[Rao Dronamraju]

Does anyone know if Q-Layer technologies are integrated into Sun's cloud
offerings presently, if not when they might be....and also is it Sun's
intention to release Q-Layer also as part of their Open Clouds
initiative?...i.e. can I down load and work with it?....

Actually that brings me to another question, if I develop something based
on Sun's cloud building blocks, will IBM continue to support it?....or do I
need to change course after putting in all the effort?....

comp...@googlegroups.com>

> read more ≫