Microsoft Battles Google, Amazon In Cloud Price War

8 Data Centers For Cloud's Toughest Jobs

Microsoft announced on Monday that it will match Amazon Web Services cloud pricing, reducing prices up to 35% on compute virtual servers and up to 65% on storage, effective May 1. Microsoft executives have previously said they planned to match AWS pricing, but that was before both Google and Amazon enacted steep drops in storage and virtual server instances.

Microsoft also plans to implement a new tier of basic virtual servers, priced 27% lower than the standard instances that will remain in step with AWS. Basic servers will have the same virtual resources as standard but without the scaling capabilities of standard servers. No load balancing or auto-scaling service will be included. The new basic servers will be available starting April 3, according to Steven Martin, general manager of Microsoft Azure.

Martin said the new basic instances are more directly comparable to Amazon Web Server general-purpose instances, which generally refers to Amazon's M-instance series. He explained the difference between standard and basic: "Basic instances will have similar performance characteristics to AWS's equivalent instances, while the Standard instances will maintain their favorable performance [over AWS].

"We recognize that economics are a primary driver for some customers adopting cloud and stand by our commitment to match prices and be best-in-class on price performance," Martin wrote. Some observers wondered whether Microsoft was going to maintain its "match-the-market-leader" stance as Google and Amazon both implemented sharp price cuts. Microsoft waited until several days after Amazon's March 26 announcement to announce its price drops.

[Want to learn more about Amazon's recent price changes? Read Amazon Counters Google Cloud Price Cuts.]

Asked where SoftLayer stood on the recent pricing moves, IBM SoftLayer CEO Lance Crosby replied in an email message, "Amazon's price cut is for only for selected offerings and it will ultimately make minimal impact on their client's monthly billing."

Amazon's price cuts include many of its most popular server instances. Its popular M3 general-purpose server, with a virtual CPU, 3.75 GB of RAM, and 4 GB of solid state disk, dropped by 38% and is available now for $.07. Cuts on older server instances, such M1 and C1 or C2, vary from 10% to 40%.

"Strictly speaking, when you include the costs that AWS and Google add on to base server price -- but that IBM SoftLayer includes standard with every server -- you see that our prices are in fact very competitive. IBM's goal is to provide the best price for performance in the market," Crosby said. IBM's "all-in pricing" includes bandwidth and support for bare metal and virtual servers as well as inbound, intra-data center and inter-data center bandwidth for data transfers at no additional charge, he added.

Crosby pointed out that Google's and Amazon's price drops are "less than meets the eye" because customers can't easily measure whether they're getting the same performance from workloads as they had previously. Asked if prices will continue to drop, Crosby replied, "One can only assume both AWS and Google are going to continue to offset the price drop by further oversubscription of their services to maintain margins." Oversubscription refers to the practice of loading too many additional virtual servers onto available hosts ito maintain previous service levels.

In effect, Crosby is saying both Google and Amazon are milking their infrastructure to host more customer VMs without increasing its capacity to maintain a pre-existing service level. "The financial benefit will be erased by the performance degradation, leaving the customer at status quo," he added.

Rackspace is another cloud service provider that could opt to match or ignore the Google/Amazon cuts. So far, the company has not announced any price changes.

IBM, Microsoft, Oracle, and SAP are fighting to become your in-memory technology provider. Do you really need the speed? Get the digital In-Memory Databases issue of InformationWeek today.