Testing The World Wide Cloud

[Reuven Cohen]

For all the discussion about the potential opportunities for the world wide cloud, there have been very few real world examples of applications that can take advantage of this idea. Today, in possibly one of the first true killer applications for multiple global cloud providers, SOASTA a provider of cloud based testing platform has announced its new CloudTest Global Platform.

SOASTA has unveiled an ambitious plan to utilize an interconnected series of rationalized cloud providers for global load and performance testing Web applications and networks. They're calling this new service the CloudTest Global Platform, which is commercially available today, and is said to enable companies of any size to simulate Web traffic and conditions by leveraging the elasticity and power of Cloud Computing.

In full disclosure, over the last several months I've gotten to know Tom Lounibos, the Co-founder and CEO of SOASTA quite well. At Enomaly, we've worked on several projects together so I can honestly say Tom is a the kind of guy a young technology entrepreneur (like myself) strives to become. Tom is a visionary and has an impressive resume to prove it. He brings more than 30 years of experience in building early stage software companies as well as leading two companies to successful IPO's. (Remember when there was an IPO market?) Most recently, Tom was CEO of Dorado Corporation, focused on Enterprise Lending Automation. In the world of cloud CEO's Tom is one to keep your eye on.

Back to what I find interesting about this new scheme; traditionally performance testing has been a kind of "best guess" scenario. Although there are many testing frameworks available most of which create a hypothetical experience using a set of static machines typically limited to one or two geographic locations. With the emergence of a global supply of regional cloud providers SOASTA is tapping into almost limitless capacity to test your application environment in a proactive fashion. Until the emergence of cloud based infrastructures testing beyond a few hundred thousand users was impossible, now you can slap together a few regionalized clouds and realistically see how 3 million or more users around the globe will actually experience your application and infrastructure. This is specially important in emerging markets such as China and India where even a low usage site can routinely get millions of users.

I think the the idea of a global testing platform is very intriguing for a number other reasons as well. Although they're pitching "CloudTest" as a testing / performance tool there is nothing saying that it can't be used as part of a proactive monitoring / scaling environment where you periodically test performance thresholds. In this proactive scaling approach you may want to predefine when, where and how you scale your infrastructure based on the real world conditions your users are "actually" experiencing. When your application is in production you could use your performance tools for proactive analysis allowing for BPM and other performance based policies to be defined ahead of time ensuring a consistent Quality of Experience.

In examining the opportunity for a world wide cloud the idea of using the quality of a user experience as the basis for scaling & managing your infrastructure will be a key metric going forward. Scaling based solely on "load" is a relic of the past. The problem is a given cloud vendor/provider may be living up to the terms of their SLA's contract language, thus rating high in quality of service, but the actual users may be very unhappy because of a poor user experience. In a lot of ways the traditional SLA is becoming somewhat meaningless in a service focused IT environment. With the emergence of global cloud computing, we have the opportunity to build an adaptive globalized infrastructure environment focused on the key metric that matters most, the end user's experience. Whether servicing an internal business unit within an enterprise or a group of customers accessing a website, ensuring an optimal experience for those users will be the reason they will keep coming back and ultimately what will define a successful business.

--
--

Reuven Cohen
CCIF Instigator
www.cloudforum.org

blog > www.elasticvapor.com

[scott radeztsky]

what a great example of thinking a few steps ahead on the chessboard.

Reuven Cohen wrote:

>
> SOASTA has unveiled an ambitious plan to utilize an interconnected
> series of rationalized cloud providers for global load and performance

> testing <http://en.wikipedia.org/wiki/Performance_testing> Web
> applications <http://en.wikipedia.org/wiki/Web_application> and

> networks. They're calling this new service the CloudTest Global
> Platform, which is commercially available today, and is said to enable
> companies of any size to simulate Web traffic

> <http://en.wikipedia.org/wiki/Web_traffic> and conditions by leveraging
> the elasticity <http://en.wikipedia.org/wiki/Elasticity_%28economics%29>
> and power of Cloud Computing <http://en.wikipedia.org/wiki/Cloud_computing>.

--
Scott Radeztsky, Ph.D.
Sun Principal Engineer
Chief Technologist, Americas Systems Engineering
Sun Microsystems, Inc.

(312) 952-6761 cell

* Customer First * Integrity * Respect for the Individual *
* Teamwork * Financial Success * Openness * Fun *

[C Wegrzyn]

Reuven, I hate to disagree. While it shows how the cloud can be used it
really doesn't show something unique to the cloud. Load testing, even to
simulate a million users could just as well have been done with a simple
set of machines. In fact with machines spread out over the globe (for
instance using the IBM datacenters) you could even stress test the network.

In my mind the real applications that will show up the cloud are the set
of applications that were discussed in the days of associative
processors. They are applications in which each processor can work
independently of one another and contribute just a small part of the
whole answer. I keep thinking of Hadoop and similar applications.

Chuck Wegrzyn

Reuven Cohen wrote:
> For all the discussion about the potential opportunities for the world
> wide cloud, there have been very few real world examples of applications
> that can take advantage of this idea. Today, in possibly one of the
> first true killer applications for multiple global cloud providers,

> SOASTA <http://www.soasta.com/> a provider of cloud based testing

> platform has announced its new CloudTest Global Platform

> <http://www.soasta.com/blog/?p=142>.

>
> SOASTA has unveiled an ambitious plan to utilize an interconnected
> series of rationalized cloud providers for global load and performance

> testing <http://en.wikipedia.org/wiki/Performance_testing> Web
> applications <http://en.wikipedia.org/wiki/Web_application> and

> networks. They're calling this new service the CloudTest Global
> Platform, which is commercially available today, and is said to enable
> companies of any size to simulate Web traffic

> <http://en.wikipedia.org/wiki/Web_traffic> and conditions by leveraging
> the elasticity <http://en.wikipedia.org/wiki/Elasticity_%28economics%29>
> and power of Cloud Computing <http://en.wikipedia.org/wiki/Cloud_computing>.
>

> www.cloudforum.org <http://www.cloudforum.org>
>
> blog > www.elasticvapor.com <http://www.elasticvapor.com>
>
>
> >

[Reuven Cohen]

Sure I could lease some IBM servers or set up a few dedicated servers in several regional co-los. but I can get 5,000 servers on 5 continents for a hour on a Saturday evening?

r/c

 

[C Wegrzyn]

So cloud computing is merely the ability to harness the power of massive
computing power on an instance? Nothing more or nothing less?

Chuck

[l...@aero.org]

If folks are really interested in doing controlled experiments,
PlanetLab or EmuLab spring to mind.

--Craig

[C Wegrzyn]

How are they available? Do you have to be a member of some sponsoring
organization?

C.

[Reuven Cohen]

On Wed, Feb 18, 2009 at 2:29 PM, <l...@aero.org> wrote:

If folks are really interested in doing controlled experiments,
PlanetLab or EmuLab spring to mind.

--Craig

 

The problem with PlanetLab, EmuLab, Open Cirrus or even Internet2 is unless your either an academic or a cloud industry insider you can't gain access very easily. We've been using Open Cirrus at Enomaly because we're in the loop. For everyone else, you're out of luck. Things like Soasta's Global Cloud Test democratize large scale global testing bringing these tools to those who couldn't either afford it or who lacked the personal connections to access it.

r/c

[l...@aero.org]

PlanetLab accounts are limited to participating organizations but with 400+
sites worldwide (and presumably that many corporate and university members),
there might be some one on this list with access.

The EmuLab at Utah is open to any non-commercial, research project,
including corporations.
We have two smaller emulabs in-house, with buffering capabilities sized for
emulation of network protocols up to geosync orbit.

--Craig