Google recently got in the IaaS cloud market to go up against Amazon, but researcher says the battle is in its early days
Google's new IaaS cloud boasts strong compute performance but lacks the breadth of features in Amazon Web Services' Elastic Compute Cloud, according to one industry analyst's side-by-side comparison of the services.
Neither company provides details of the silicon chips within its servers, but analyst Chris Gaun from Ideas International (recently acquired by Gartner) has used information in public statements to determine the hardware behind each vendor's cloud. Google has said it uses Intel Sandy Bridge processors and that each unit of its Compute Engine delivers performance matching that of at least a 1.0- to 1.2-GHz 2007 Opteron chip. Other media have reported that Google uses 2.6-GHz processors, which leads Gaun to believe the company has Xeon E5-2670 chips, the only ones on the market at the time of Google's announcement that deliver that level of raw compute power.
Gaun believes Google is running the high-capacity chip across its cloud infrastructure, while Amazon makes it available in certain instance types for Elastic Compute Cloud customers, including in its recently announced high I/O extra large cluster compute offering. "Google seems to be running only the latest and greatest chips on the market, while Amazon has a wide variety of chips for customers to use," Gaun says.
Amazon isn't standing pat either. AWS on Wednesday, for example, announced the ability to set the input/output operations per second (IOPS) in Elastic Block Storage.
There are other differences between Google Compute Engine, which is still in limited preview mode, and Amazon cloud compute services, which the company launched in 2006 and made generally available in 2008. AWS has 14 different sizes of compute instances, ranging from small virtual machines with 1.7GB of memory, to extra-large compute clusters with 60.5GB of memory, whereas Google has only four. Google also makes the fiber-optic links between its own data centers available to cloud customers. AWS has a variety of accommodating features in its cloud though, such as the EBS volumes, relational database services, load balancers and others.
The two companies are appealing to different customers, Gaun says. While AWS is targeting technology-reliant businesses that are turning to the cloud to host their websites, databases and storage, Google is focused initially on research and development teams that may have a need for high-performance computing to complete a project, for example. The strategy is seen in the pricing models: AWS offers reserved instance pricing discounts, in which customers agree to use a compute instance for months or even years. Google's cloud is priced by smaller time chunks and therefore aimed at shorter-lived projects.
Gaun says if Google wants to compete in a broader market with Amazon, it will likely have to offer a discounted pricing option for long-term use. That may come in time, Gaun predicts, given that the company's cloud computing offering isn't even generally available yet.