Cloud Computing

Cloud Computing is generating a lot of smoke, but it is hard to find flammable material that makes it “hot”. I believe that Cloud Computing is unlikely to become Essential Service, but will be more like Public Transport – cheap, mass market, limited function and “acceptable” for some.

Note

This article was originally published in November 2008. I was rereading it and realised that nothing has changed in the last 14 months of “Cloud Computing” and it’s worth saying it again.

Introduction

With the level of hysteria on Cloud Computing reaching astonishing levels, I find the huge claims by major league pundits ring hollow but now that management teams are issuing requests that are flowing down into the Technical Design process it seems time to reflect on what Cloud Computing will do to the design process. Cloud Computing is something that we have been doing for years in our data centres, the only difference is that it looks simpler.

The marketing message is beguiling and quite false. Here is what I am hearing: “so you simply move your applications in to the cloud. The cloud will deliver compute and storage resources to deliver whatever, wherever you need it. Moreover, it will cheaper than you are paying.”

What rubbish.

The Seduction

The really seductive part of this pitch, is that all your “pain” around managing an IT infrastructure will go away. You can almost hear the senior management wishing on stars – “I can choose to send all those cost centres to an outsourcing company” with a huge smile on their faces. You know they are thinking “Less head count, less hassle, and better cash flow”.

It’s a good message, and it’s going to work for a lot of companies who are going to waste spend a lot of time on looking at Cloud services.

The Reality is not fully integrated

The only difference between the “Cloud” and the current environment in our data centres is the level of integration between all the components. The seduction stems from the idea that all the storage, all the compute resources and networking connections are already in place. This is possible because they resources are shared around.

From reading about the genesis of Amazon, the service was started after a successful IT project got the storage / compute / network resources into a coherent operational platform. Pretty simple really. To summarise, the Amazon people all got together and agreed that they need to automate the setup of compute / storage / network resources to support the massive rollout that is needed.

So they worked out ways to allow the network / server / storage to automate the creation / management / deletion of virtual instances. As a bonus, each virtual instance is highly available and can move around the cloud as needed. However, this means that your choices for operating platforms are quite limited and don’t allow for customisation aka default builds are mandatory.

The Kicker

There is the downside of the cloud, you get what your given, and a substantial about of the computing platform will be proscribed by the provider. As Henry Ford said, “you can have any colour you like, as long as its black”.

Does this work for your environment ?

Where it will fail and fail early.

This website is hosted with Media Temple. It uses a shared database server and shared web server to host the content – it saves money to share resources, right ?

A recent experience was that during the day, performance of the web site get very poor, with response times of up 60 seconds. The tech support people claim that it is not their systems. Yet, I have scripts that seem to prove that the database server is not performing well and claim that my software is the problem.

This has been going on for some days now and there appears to be no resolution. The provider claims that their systems are performing as expected ((does that mean good, I will never know)), and that nothing is wrong. I know that performance is poor, and can see it.

It won’t take too much of this to give Cloud Computing a bad name. Something very similar happened to Application Service Providers back in 1999/2000. At this time, Citrix was a popular way to outsource Microsoft Office and Exchange servers thus centralising the computers and expertise. Telcos and service providers were quick to offer these services, but they didn’t last very long before customers went back to the traditional model. An absolute shedload of cash was spent before it finally petered out.

The main problem was agreeing on what is the problem. Can you application live with this ? Can your management live with you saying “It ain’t me, its the cloud”.

Cloud Services that are succeeding – there are some.

This is reasonably straight forward – any service that is a simple commodity will be readily transferred to the cloud. Thus services such as email, word processing, simple image processing are existing examples.

In this view, this is taking existing services such as Google Mail and attaching the term ‘cloud computing’ to it. But this definition any version of Microsoft Exchange is also ‘Cloud Computing’ service, and its an internal Cloud.

These services are and will be successful because they are simple, time insensitive and not mission critical ((and before you email me, corporate is NOT mission critical, its can be out of service for an hour or more before the world starts to end, that’s NOT critical, just mission-sensitive)).

The Cloud will be what you are given – not what you want!

When the Cloud providers finally get established, they will have to provision repeatable, scriptable, automated, standardised platforms. However you cut it, this is unavoidable because it is unimaginably difficult to support hundreds of differing configurations and varietal systems. As Henry Ford said in the early days of mass production “Any customer can have a car painted any colour that he wants so long as it is black.”

As long as you want these “predefined” services, then Cloud Computing is going to fit you just right.

Isn’t Public Transport just “Great”

As long a you can accept the idea of the service being less than you might want, then Cloud Computing might work. So just like Public Transport, it works well so long as the train station is at the end of your road, its cheaper than the alternatives. It isn’t very comfortable, and when you get off the bus, you have to walk a bit to get to your destination.

Therefore most companies will build their own Private Clouds because these systems can be tailored to their requirements. While Public Clouds will be limited in flexibility, features and capability.

I happily use an email “cloud computing” service ((Google Gmail)), because I can live with the price and lack of features. In this paradigm, Cloud Computing is for poor people, the not well off, badly advised and no other choices.

When it comes to delivering my online web strategy for a corporate company…. well, public transport works for some people who have no other choice but it doesn’t work so well for most people. Private transport takes you where you want to go and back again.