Exploits & Vulnerabilities
Apple's iCloud - an Ambient Cloud from the Future
Cloud computing has a big capacity problem. No matter how many data centers we build, we’ll always be behind the curve of demand for capacity. I predict that ambient clouds will become the dominant way of delivering cost-effective cloud capacity.
Cloud computing has a capacity problem. A big one. No matter how many data centers we build, we’ll always be behind the curve of demand for capacity. I’ve been predicting since 2008 that ambient clouds will become the dominant way of delivering cost-effective, performant cloud capacity. It’s the only way I can imagine serving every human on the planet with cloud computing without exhausting our electrical generation capacity for data centers. In fact, one of the reasons I’m working for Trend Micro is that we have our own sizeable ambient cloud that handles more than 5 billion transactions a day with a fraction of the data center footprint.
Ambient clouds are also at the center of the competitive strategic landscape in IT companies right now. This post explains how to use ambient clouds as a way to view the strategies of Microsoft, Google, Apple, and others. With Apple’s new iCloud architecture, they’ve shifted the landscape from my admittedly qualitative perspective. And they did it with the most advanced ambient cloud architecture I’ve come across.
At first blush, you might think I’m nuts when I say it’s an ambient architecture. As Rich Miller at Data Center Knowledge has reported , the iDataCenter:
• will cover about 500,000 square feet — about five times the size of the largest shared data center we ever built when I was at Exodus Communications, the first big collocation company. And Apple’s iDataCenter cost around $1 billion to build, about 2x what Google and Microsoft spend on a data center. Then they built it using high availability tools, including clustering, and using technology from IBM, Veritas and Oracle, exactly the opposite of Google’s (and most IaaS provider’s) focus on commodity storage and servers. Then to top it off, Apple ordered 12 petabytes of Isilon file storage from EMC.
But that’s exactly why this is a massive scale ambient cloud. The centralized command and control portion of an ambient cloud HAS to be high availability. Some parts – core data and system-of-record data belong on HA systems. They had better not suffer from data loss, like Amazon did in their unfortunate 36 hour recent outage. If people will use iCloud for syncing everything — photos, audio, documents and more — across all their iDevices, they have to believe that their data is safe and won’t disappear in an outage. The operational stakes are very high here – if Apple’s customers choose to store their family photos in Apple’s iCloud, an outage with data loss can be a company-credibility-ending event.
But the reason that this is the most advanced ambient cloud is that Apple must have done some math and realized how much it would cost to store everyone’s music and video in their cloud. No matter how many cheap disks you aggregate, you still run into electricity and space costs. Instead, Apple seems to be saving that massive storage capacity for non-music and video files. Unlike Google, Apple is going to store individual copies of everybody’s media files on their iDevices instead of in the cloud, and use the HA centralized command and control to sync copies of newly purchased music. It’s massively more efficient than storing copies in the cloud. Even Apple’s impressive 12 petabytes would be overwhelmed with the contents of every iPod in use today.
The new move gives Apple a cost advantage over Google, and Apple has a much larger ambient cloud footprint than Google, at least today. As my first ambient cloud post suggests, Google is working quickly to make Android and Chrome into their ambient cloud platform. I’m in the process of revising my ambient cloud rankings to reflect this change. I think Apple’s simple, clean design ethic just got applied to their data center and architecture, which is not what I would have expected.