Rackspace CTO John Engates has taken a big whack at Amazon Web Services, accusing the rival platform of purposely confusing the meaning of "dedicated," "server," and even "cost.” Amazon announced July 10 that it would drop prices between 37 percent and 80 percent for chunks of virtual-server compute power called "dedicated instances." That could pose a direct threat to Rackspace, which offers dedicated servers and infrastructure as part of its hybrid cloud services. In a blog posting, Engates suggested that the "dedicated" EC2 services whose prices Amazon cut are only vaguely similar to actual physical servers dedicated full-time to a single customer, which Rackspace offers at one end of its spectrum of hosted services. Amazon cut at fees for Dedicated On-Demand Instances and Dedicated Reserve Instances, as well as the additional fee for having a Dedicated Instance running in a specific region. Specific definitions of many cloud-computing services vary depending on the provider describing them. On EC2, an "instance" is roughly equivalent to a virtual server—which, by default, runs on a physical server that also runs virtual-server instances belonging to other customers. While multi-tenant hosting is the basic assumption for most cloud services, many companies shy away from it to avoid running apps with sensitive data on shared hardware, or to keep control of security and performance of complex or demanding applications. Many customers want a service provider to host an actual physical server stripped to the bare metal, so they can load and configure the software they want, and control everything on it as if it were sitting in their own datacenters. Big customers typically want a range of cloud services ranging from lightweight apps on shared, low-cost multitenant hardware to virtual private clouds made up of physical servers connected directly to the customer's datacenter—all while not touching a shred of any other company's cloud at all. "The public cloud is a powerful technology, but it isn’t the answer for every business or every workload," Engates wrote in his posting. "The real conflict lies in the way that AWS defines dedicated computing, which is at odds with the view of the rest of the industry, including Rackspace." Dedicated instances may run on their own servers, but those servers live within Amazon's cloud and can connect to the customer only through AWS. They also run only as virtual machines, which come with performance overhead that requires customers to add I/O or CPU resources to overcome. Rackspace customers running bare-metal servers can get rid of the abstraction layers and run their workloads directly on a physical server they control completely, Engates added. Rackspace's definition of "isolation" also allows customers to choose storage hardware that is equally dedicated and isolated, and connect virtual private clouds only through private network links to the customer's own datacenter. Customers can also choose to house their hardware in facilities separate from those used by other customers, and rely on their own separate supplies of power as well. "When using EC2 dedicated instances, you are also on your own hardware, but still connected to Amazon’s public cloud. You’re just a dedicated, single-tenant slice of that cloud. If the AWS public cloud suffers an outage, you will be affected," Engates wrote. Rackspace customers can also choose the actual hardware they run on, rather than accept the commodity level servers provided by Amazon. Though he's partisan toward Rackspace as its senior executive, Engates’ differing definitions of both dedication and isolation are actually valid, according to Wells Fargo financial analyst Gray Powell. "Or contacts suggest customers still cannot fully optimize the environment to the workload as one can with a dedicated server in either a managed hosting or colocation environment," he wrote in a bulletin circulated to Wells Fargo customers, as quoted by Barron's. Amazon is likely to get a boost from both the more-dedicated version of its cloud service and from customers who choose it due to its size and capacity, wrote Needham & Co.'s Richard Kugele in a similar bulletin. That size gives AWS a tremendous competitive advantage over other cloud providers, even if "the fact that AWS' reliability issues keep its addressable market limited to low-end/midrange, non-mission-critical applications," Kugele wrote. Amazon launched its dedicated services to overcome issues about reliability and control, of course. But the attraction of Amazon goes beyond just size and celebrity, according to Amir Salihefendic, founder of Web services Doist, Todoist, and Wedoist. Though Doist signed with Rackspace for its hosting expertise, it switched to AWS partly because customers in more remote countries couldn't always reach the service. With AWS, that problem disappeared, as did the desire to run exclusively on private networks and hardware. Shared or semi-private clouds make it far easier to scale up or down to match demand without requiring Doist to maintain data and servers on its own in multiple datacenters around the world. Image: rickyd/Shutterstock.com Rackspace CTO John Engates has taken a big whack at Amazon Web Services, accusing the rival platform of purposely confusing the meaning of "dedicated," "server," and even "cost.” Amazon announced July 10 that it would drop prices between 37 percent and 80 percent for chunks of virtual-server compute power called "dedicated instances." That could pose a direct threat to Rackspace, which offers dedicated servers and infrastructure as part of its hybrid cloud services. In a blog posting, Engates suggested that the "dedicated" EC2 services whose prices Amazon cut are only vaguely similar to actual physical servers dedicated full-time to a single customer, which Rackspace offers at one end of its spectrum of hosted services. Amazon cut at fees for Dedicated On-Demand Instances and Dedicated Reserve Instances, as well as the additional fee for having a Dedicated Instance running in a specific region. Specific definitions of many cloud-computing services vary depending on the provider describing them. On EC2, an "instance" is roughly equivalent to a virtual server—which, by default, runs on a physical server that also runs virtual-server instances belonging to other customers. While multi-tenant hosting is the basic assumption for most cloud services, many companies shy away from it to avoid running apps with sensitive data on shared hardware, or to keep control of security and performance of complex or demanding applications. Many customers want a service provider to host an actual physical server stripped to the bare metal, so they can load and configure the software they want, and control everything on it as if it were sitting in their own datacenters. Big customers typically want a range of cloud services ranging from lightweight apps on shared, low-cost multitenant hardware to virtual private clouds made up of physical servers connected directly to the customer's datacenter—all while not touching a shred of any other company's cloud at all. "The public cloud is a powerful technology, but it isn’t the answer for every business or every workload," Engates wrote in his posting. "The real conflict lies in the way that AWS defines dedicated computing, which is at odds with the view of the rest of the industry, including Rackspace." Dedicated instances may run on their own servers, but those servers live within Amazon's cloud and can connect to the customer only through AWS. They also run only as virtual machines, which come with performance overhead that requires customers to add I/O or CPU resources to overcome. Rackspace customers running bare-metal servers can get rid of the abstraction layers and run their workloads directly on a physical server they control completely, Engates added. Rackspace's definition of "isolation" also allows customers to choose storage hardware that is equally dedicated and isolated, and connect virtual private clouds only through private network links to the customer's own datacenter. Customers can also choose to house their hardware in facilities separate from those used by other customers, and rely on their own separate supplies of power as well. "When using EC2 dedicated instances, you are also on your own hardware, but still connected to Amazon’s public cloud. You’re just a dedicated, single-tenant slice of that cloud. If the AWS public cloud suffers an outage, you will be affected," Engates wrote. Rackspace customers can also choose the actual hardware they run on, rather than accept the commodity level servers provided by Amazon. Though he's partisan toward Rackspace as its senior executive, Engates’ differing definitions of both dedication and isolation are actually valid, according to Wells Fargo financial analyst Gray Powell. "Or contacts suggest customers still cannot fully optimize the environment to the workload as one can with a dedicated server in either a managed hosting or colocation environment," he wrote in a bulletin circulated to Wells Fargo customers, as quoted by Barron's. Amazon is likely to get a boost from both the more-dedicated version of its cloud service and from customers who choose it due to its size and capacity, wrote Needham & Co.'s Richard Kugele in a similar bulletin. That size gives AWS a tremendous competitive advantage over other cloud providers, even if "the fact that AWS' reliability issues keep its addressable market limited to low-end/midrange, non-mission-critical applications," Kugele wrote. Amazon launched its dedicated services to overcome issues about reliability and control, of course. But the attraction of Amazon goes beyond just size and celebrity, according to Amir Salihefendic, founder of Web services Doist, Todoist, and Wedoist. Though Doist signed with Rackspace for its hosting expertise, it switched to AWS partly because customers in more remote countries couldn't always reach the service. With AWS, that problem disappeared, as did the desire to run exclusively on private networks and hardware. Shared or semi-private clouds make it far easier to scale up or down to match demand without requiring Doist to maintain data and servers on its own in multiple datacenters around the world. Image: rickyd/Shutterstock.com 