Zynga’s approach turns conventional cloud computing thinking on its head. The most frequently mentioned case for the hybrid approach of blending private and public clouds is “cloud bursting,” where the in-house data center runs the bulk of the load, spots a spike building, and offloads that extra work to the public cloud. Zynga does the opposite, launching games in the cloud when demand is lowest–as is predictability.

I loved this approach – unconventional, effective, and ultimately – cost-effective. Let’s examine the facts:

1. Cloud computing today on the public cloud is limited to dev/test scenarios – mainly because of the prohibitive costs of running a cloud operation 24X7. It’s cheaper to own your datacenter in the long run.

2. Hybrid scenarios typically talk about “burst capacity” – the ability to meet increased demand with rapid provisioning ; the issue here is that most companies except startups have invested in DC capacity and feel comfortable in handling periodic spikes with current investments, without introducing the cloud providers and adding another dimension of complexity of management.

3. AWS failure – yes, bad stuff happens. It happens in enterprise data centers too – all the time, except its not on the front page of the media. What’s that saying about a tree that falls in the forest makes no noise?

The key takeaways for me were: Align cloud usage to business usage

If a product launch is pending, the business unknowns are the most high; use cloud to mitigate that risk. When a demand pattern emerges, move it in-house to save costs.

* Do not make assumptions on an external providers architecture; due diligence is key to understand the weaknesses

* Cloud strategy needs to have a clear overarching architecture that owns both private and public parts, and the interaction thereof; the ability to move workloads inhouse is key.

* Do not skimp on Disaster Recovery plans. Understand that geo-location of datacenters is a key factor no matter how its spun (zones, areas, etc etc)

My view is that this is not an isolated incident – Data centers, whether owned or rented, on-premise or hosted or on-cloud, all fail. In fact, that’s the reason its four 9′s or five 9′s – never 100%!

As an enterprise, remember that the job of the CIO is to anticipate this risk of technology failure and plan for it. Disaster recovery is not new. So why is it that in this case companies that are complaining about AWS downtime are not talking about their DR plans? How come we dont have ONE company that is able to tout that despite AWS being their primary cloud provider, their operations continued to work without a hitch?

This is not a technical failure, its a business failure. It’s a belief that AWS is so big and so very well engineered that it will never go down. But this black swan incident is just first of many.

So if I were you, I’d go back to first principles – dust off the DR manual and look deeper into providing redundancy. One provider is NOT enough. Even in the cloud with all its supposed magic, it will take two to tango.

I used to think that all SaaS vendors should be/are offering the monthly payment option but it turns out that this isn’t the case for some very understandable reasons

My Take:

The SaaS business was something that I used to like before. From a customer standpoint, it was freaking great: pay-as-you-go, no long term committments, instant on, no need for IT etc etc. The perfect disruption to bid ERP. Something that the small guy now has access to that traditionally had a barrier of entry with upfront capital investments and an implementation burden. All that is great. Now for the negatives.

From the vendor’s perspective, cash is the fuel of any software company and SaaS vendors need a lot of it. Remember, they must acquire everything needed for successful data centers. That’s a cost on-premise vendors don’t have. Second, many SaaS vendors aren’t collecting big upfront payments. They may have customers using only a few licenses during the implementation (when a vendor’s support costs are highest and revenues the lowest). They won’t get a big up-front payment from a customer unless they price their software accordingly. They also have the usual software costs like development, sales, marketing and support to fund as well. In all, SaaS companies aren’t exactly cash-flow generating machines except for those companies with lots and lots of paying customers.

What this fundamentally means is two things:

1. Software development is similar to construction of a skyscraper, and customers are tenants in that skyscraper. The blueprint, design and construction are critical upfront costs. The fact that one vendor asks for payment upfront v/s pay-as-you-go does not make these costs go away.

In fact, the pay-as-you-go model hurts cash flow for the vendor unless, as noted above, you have tons of paying customers – in which case the business model resembles more like the telecom industry. Microsoft for example, has a policy for protecting themselves f0r 12 months which is smart. If you dont cancel within 30 days, you pay the equivalent of early termination fee or lease breakage for the rest of the 12 month term.

2. Price ceiling – this is what brings me to my title – is being set by salesforce.com as the SaaS leader.

Try exceeding this and watch how customers react. See if you can charge $85 per user per month (more than $65 that SFDC charges) for ANYTHING. This in effect, is highly limiting from a profitability perspective in any industry. So why is this important? Risk. You do not want to see a SaaS vendor go under. As long as the hype around cloud continues and easy capital is available, we will not even notice this cash flow issue. But am I the only one thinking that this bubble is going to burst and real business metrics are going to become important, again?

Net-Net: SaaS vendors are between a rock and a hard place; as the cloud space evolves, we will see alternative pricing models emerge and failures due to ill thought out profitability plans – the oligopoly on the ERP software vendors is bound to be inevtiably replicated here as well. So, choose wisely.

I’m reacting to a poll being conducted by Dell and Intel on Virtualization benefits on LinkedIN. Amongst the various slices of data, I found this chart “by Age” the most interesting and telling:

It appears that 18-24 year olds care more about High availability (50%, in yellow) than those over 55+ who care only 15%.

On the other hand, 55+ year olds care about “lower expenses” at a whopping 62% compared to only 20% for the 18-24 year olds.

This is intuitive.Folks who carry a mortgage and have children in college worry more about expenses than a 20 year old frustrated with having no access to Facebook at work or having systems fail.

Could it be that IT priorities are as much driven by demographics and psychographics as much as by title and company size?

"By 2014, we expect this to exceed $95 billion, or about 11 percent of total worldwide SMB ICT spending – indicative of a compound annual growth rate of 13 percent," Deepinder Sahni, AMI’s senior vice president for global sizing and segmentation said in a statement.

# of Azure customers: 10,000

%Apps as replacement : 50%

Pricing:

• Compute time is priced at 12 cents/hour,
• storage at 15 cents per GB stored and
• storage transactions at 1 cent per 10K.

More details on pricing at:

http://www.microsoft.com/windowsazure/offers/popup/popup.aspx?lang=en&locale=en-US&offer=MS-AZR-0003P

So per my calculations, a basic extranet to support 50 suppliers running 24X7 will cost about $10K a year. Not bad!

For one, Microsoft is “out of the closet” with their Azure and BPOS – pronounced bee-pos (Microsoft Cloud) offerings.

Also, I’ve switched employers (if you really want to know, I’m with WinWire Technologies – a solution provider in the Collaboration and Analytics space) to emphasize primarily on MS, so expect future updates to reflect that knowledge.

There has been an improvement in the general understanding of maturity of the cloud, with formal cloud benchmarks being published to help with shopping.

Salesforce.com introduced Chatter – a sort of Facebook-for the-Enterprise tool.

Amazon is continuing down the path of opening additional datacenters around the world proving cheap infra-on-demand, and offering upto $100,000 for startups that are using AWS to solve interesting problems.

But the single biggest disruption in all of this is the availability of Exchange, SharePoint, IM and Live Meeting for 10 bucks a month. I’m yet to find a customer that has economics more compelling than that price for collaboration. And I’m bullish about Microsoft this time – they have something that’s solid and will change the game for sure in ways that we have not been able to articulate yet.

If you’re getting the sense that moving mission-critical applications to the cloud is going to be more of a journey than an event, you’re probably right. No doubt this process will take years to complete, and once it does, IT organizations will find themselves playing the role of integrators of services.

And he’s kind enough to quote me:

In 2010, we’re in the early stages of moving mission-critical applications to the cloud… But before that happens, IT organizations first need to examine their application portfolio. After years of economic growth followed by a sharp downturn, companies still have a lot of redundant applications and paying monthly fees for redundant applications doesn’t make sense.

… IT organizations need to start “decomposing” their applications to determine the characteristics of those workloads.

• Numerical processing intensive
• Request/response
• Event-driven
• High concurrency/high throughput
• Ubiquitous user
• I/O-intensive

My take:

Workload view, while very appropriate,  is one of the many other dimensions to consider. The others are:

• Domain view (Business process, Apps,  Infrastructure)
• Industry (Banking, Financial Services, Insurance v/s Manufacturing v/s Life Sciences)
• Maturity

The underpinning of all cloud migrations is the desire towards providing a variable cost structure in a cost constrained economy. Towards this, it’s important not to oversimplify from a technology standpoint but really step back and assess what the business needs are, and how best can an efficient IT organization can provide this – balancing the existing infra investments with the need to provision the new.

An all too familiar scenario for those of us on the marketing side of technology companies 

Hat tip:http://technobabble2dot0.wordpress.com/2010/01/29/downfall-gartner-mq/