One of the best takeaways from Tuesday’s Writing Code for Many Clouds session was Sam Ramji (of Sonoa Systems) ’s ten tips for a platform agnostic approach to cloud development (based on his experiences creating apigee):

1. Get comfortable with virtualization

This means both your developers and your admins.

2. Limit dependency on provider specific APIs

This is how you avoid lock-in.

3. Refuse to use unique services

In other words, use abstraction layers like RightScale or even better, the open source libcloud – so you don’t have to write infrastructure level handles yourself, leaving your developers free to focus on building your application.

4. There are no SLAs on your provider’s APIs.

Build that into your design. Handle failure gracefully.

5. Understand the different strengths of different clouds.

For example, Amazon may be better for scalability, or Rightscale may be better for control over networking topology.

6. Anticipate your needs for data replication and design accordingly

You’re putting data all in one place – a worldwide singleton of data – but what about if you’re using multiple clouds, what goes where?

7. Pay attention to network topology and networking requirements

For example, you can’t do multicasting with amazon, so apigee used another product called VPNcubed – giving service across Rackspace and Amazon. This is closely related to the data replication consideration.

8. Consider the granularity of the requests that you will load balance

API level granularity makes your ability to run on multiple clouds much greater

9. You’ll still overbuy compute power

You still have to tell Amazon ahead of your need. But the waste ratio is less, and the racking and stacking problem is gone.

10. Monitor the monitors!

There are no SLAs on your APIs so you should have at least one box to monitor your monitor.

Another highlight of the Cloud Development track was Jinesh Varia (of Amazon Web Services) ’s excellent session on Design Patterns for Cloud Architecture. He has written a whitepaper on this, and his slides will be available soon.

The first set of keynotes at the first ever Cloud Connect conference took place this morning, two fast-paced sessions with thirteen presenters having ten minutes each. Here’s a summary of ten takeaway messages from the the event (including the interesting stats on the cloud adoption shown right):

1) Cloud computing is the grey matter of our distributed nervous system.

Alistair Croll (Bitcurrent) opened with a look at “clouds and crowds”. He explained that cloud computing is just the next step in a timescale of human evolution that began with farming and irrigation and brought us the industrial revolution and electricity in every home. Now that we are sharing more and more and living online, cloud computing is the enabler for a distributed human consciousness.

2) Cloud computing accelerates the supra-national evolution of the Internet.

Read the rest of this entry »

In a few weeks, we’ll hold the inaugural Cloud Connect in Santa Clara, California. It’s actually the continuation of a series of events David Berlind launched around cloud computing, plus a spinoff of last year’s Enterprise Cloud Summit, plus a bunch of new content.

We’re pretty excited, because this is the first time Bitcurrent has helped build an event from scratch (unless you count Bitnorth, that is, but Cloud Connect is a beast of a different magnitude.) There are four days of content, built around three audiences: those who buy and finance cloud decisions; those who build cloud applications, and those who have to run the cloud platforms.

Getting here has been an interesting experience. Here’s what we did, plus an easter egg for reading all the way to the end. Read the rest of this entry »

After Cloud Connect in Mountain View, Barton George asked me some questions on cloud computing and some 2009 predictions. Here’s the clip; his writeup is available on his blog.

I’d like to compliment Barton on his ability to pick exactly the worst moment in the entire video clip to use as a thumbnail. And yes, I am this dorky looking.

I’m never big on predictions; they’re an easy way to look dumb in retrospect and few people keep score on accuracy after the fact. But it does seem to be a January ritual.