Basic Idea Behind the Cloud

tummy_developer

unread,

Apr 16, 2009, 3:45:34 AM4/16/09

to BigCloud

Hi People,

This Tummy Developer (Thomas) from chennai. Wants to know about the
Basics of Cloud ! Instead of the Current hosting types how clouds
going to be the alternative and cost effective ?

Regards,
Tummy Developer.

Saifi Khan

unread,

Apr 24, 2009, 7:34:26 PM4/24/09

to BigCloud

Hi Thomas:

Let's say you want to make a online tax filing service.

You'd typically develop the application (server-side) and
web UI (client-side) in some language of your choice.

At some point in time, you'll want it to be accessible to a
large number of people. One easy way, is it to do virtual
hosting on web (ie. internet).

You pay some hosting fees monthly and go about running your
business.

This is the application service or service on internet that
most of us have used. Read on ...

To deploy your service you have certain resource available to
you:
. disk space
. memory
. processor
. bandwidth

Let's say a lot of people are using your service to check out
their tax filing status, so you are doing a lot of computation
and choking up, even though the bandwidth is not exhausted.

What do you do ? How do you handle the demand ?

Common approach was to load balance the service and host it on
multiple boxes. Nice the system is able to take the load, but
you just spent a lot of money acquiring and maintaing the
additional box. Accountants(i mean CFO) like to call it CAPEX
(capital expenditure).

A better approach would to be do what is done typically in a
marriage - horse of hire, priest on hire, utensils on
hire, food on hire, lighting on hire, chair on hire, smile on
hire, band on hire, cook on hire etc etc. to provision for the
large number of people for a day or two.

After the marriage - we're back to our own 800, plate, food,
tubelight, chair, plates, frown, mp3 player, kitchen etc.

Additionally, during the entire exercise, if 4 more helpers are
needed, it is the responsibility of the service provider to
figure it out. For us, its transparent or 'virtualized'.

So, we goto Baba Cloud service and evaluate their 'monsoon' package.
In case we don't like it, we goto Pappu Cloud engine and
evaluate their 'barsat' package :) you may see pythons there :)

Once the rental deal is worked out, we signup.

All we need is our application/service and credit card !

Accountants(i mean CFO) like to call it OpEx (operational
expenditure). This is manageable, explainable on a balance
sheet. OpEx is marketable and VCs also like this.

To summarize, resource stretch is provisioned by projecting it
to the service provider on hire. The hiring costs are much lower
since the service provider provides the same service to multiple
clients and thus costs are spread. The access models are built
on virtualization (ie. processor, disk, memory, network).

From a business perspective this is a 'bikau' solution if not a
very 'tikau' solution but "chalega" (aka eventually consistent).

As a service creator, your "entry" costs are low and you can
board the service (SO) tractor asap. If need be, you attach a
trolley for the extra load :) and pay a small additional rent.

In the long run, as more service creators, ride the tractor, it
will fundamentally alter the behaviour and approach to
application deployment, ie. no more SUVs (aka my dedicated
guzzler server).

Actors:
. munna (end-user)
. you (service creator)
. baba (cloud service front end1)
. pappu (cloud service front end2)
. gabbar (cloud service provider)
. sambha (provisioner at cloud service provider)
. thakur (compliance for cloud)
. basanti(cloud control panel)

Please see
http://www.ibm.com/developerworks/websphere/techjournal/0904_amrhein/0904_amrhein.html

Hope this helps.

thanks
Saifi.