Lessons From AWS Part I: The Crush of the Boa Constrictor
The cloud industry is often portrayed as a race between Amazon’s AWS, Google’s Compute Engine and Microsoft’s Windows Azure. The reality however, at least to date, is more like AWS and the also-rans. The lesson is scale and the classroom is Walmart.
For all the recent announcements out of Google and Microsoft, both still come up short of AWS in terms of available features and ecosystem. In fact, AWS not only offers products that match all of the most recent features that became generally available from its two competitors, it surpasses them by a significant margin. Moreover, while Google and Microsoft have some third-party companies that provide additional services on top of their cloud, neither is a match for the sprawling third-party ecosystem surrounding AWS.
Google’s strength in networking owing to its global fiber footprint and Microsoft’s SSD-powered storage capabilities are formidable. So are both companies’ cash hoards. But they will not be enough to catch AWS. The reason is this: while most customers are using AWS for its basic compute and storage services, as more enterprises migrate more workloads to the cloud, they will likely want to buy as much capacity from a single vendor as possible. There are numerous motivations for this, ranging from cost and integration to security and governance. This gives AWS the type of insurmountable advantage that Walmart still commands.
Is there a Target to AWS’s Walmart?
In the 1980s, Walmart invested more heavily in technology than any of its competitors. This gave the company overwhelming advantages in warehousing and distribution. As it learned more about its customers, Walmart extended its advantages into sourcing and merchandising. The company then plowed the cost advantages into discount pricing that no other retailer could match. In fact, there were instances where Walmart could sell certain products below what it cost some of its competitors to purchase the same product.
The result was a sad story in retail road kill. As Walmart opened new stores, flocks of discounters went out of business, while others merged unsuccessfully in attempts to reduce costs. When Walmart diversified into the supercenter format, scores of supermarket and drug chains felt similar pain.
Target was the only retailer with comparable scale. It succeeded by maintaining a tight focus on quality and fashion in order to differentiate itself from the behemoth. It followed this formula into the supercenters as well.
The point in this is that in a commoditized business where scale is critical, the key to competing against the 800-pound gorilla is differentiation. And in the cloud, that is going to be on quality of service, service offering (variety or geographic) or platform (read: OpenStack, but more on this in Part II). Joyent’s recently announced price cuts to and server slicing to match up with AWS rings hollow and is likely to end in tears. To its credit, Joyent is also offering a few dozen configurations that are not available on AWS – yet.
But over time, trying to compete with AWS on pricing is like trying to wrestle with a boa constrictor. With four data centers in total (three in the U.S. and one in Amsterdam), Joyent is going to get crushed. Its plans for additional data centers in other regions will not give it the scale it needs.
And judging by the two most recent quarterly results reported by Rackspace, they too will suffer the same fate. On its conference call, the VP of finance said, “Don’t be surprised if we are continually lowering prices on certain products. We are a cost-plus shop.” That sentence could come back to haunt Rackspace in a big way. For it is insane for cloud providers to try to compete on raw compute pricing with AWS.
Cloud providers can either sell against AWS’s outages with better reliability or they can try to distinguish on service offering. In the latter, it will probably require some specialization. One example would be allowing customers to scale up processor threads, memory, disk and flash independent of each other depending on the workload needs.
That type of flexibility would require excess infrastructure and beget greater management complexity. But offering customers dynamic configurations would represent a value-added service that could command a premium price, especially for workloads that do not scale well horizontally across virtualized server slices as they do in a larger virtual footprint.
This would be similar to what the category-killers such as Staples, AutoZone, PetSmart or Sports Authority did. Focus on one of Walmart’s aisles and try to out-execute the giant on either selection or service. For Google and Microsoft, the alternative is the Target strategy. They both have the resources to make a go of it – even as the number 2 and number 3 players. The others can only hope they won’t end up as Kmart, which was later merged into Sears.
