Google Outsmarts Competition With Custom Machine Types In The Cloud

Customers migrating to the public cloud get overwhelmed with the number of instance types and the categories offered by the providers. As of November 1^st, 2015, Amazon EC2 has 39 instance types to choose from 6 different instance families. Though AWS takes pride in rapidly adding features and services to its platform, it hasn’t done much in providing appropriate guidance in choosing the right instance types. Amazon continues to add at least two new instance types every year. Jeff Barr, the chief evangelist and the official blogger at AWS, published a detailed history of EC2 instances.

Microsoft Azure is no different in overwhelming the customers. It has over 30 VM types with varied capabilities and pricing options but with very little guidance on choosing the right configuration.

Like most of its competitors, Google has a choice of 18 predefined machine types that offer varying power of CPU, memory and storage across 4 categories. Each machine type has its own pricing and is billed separately. For example, customers running transactional databases choose from high-memory machine types while running the development and test workloads on standard machine types.

The challenge comes when customers try to map their on-premises server configuration with the available machine types. In most of the cases, they settle for a configuration that’s only a close match but not a symmetrical match. That’s where Google Compute Engine custom machine types come into the play. Customers can create a machine type with the number of cores and memory that mimics their existing server configuration. This feature works great for workloads that need raw computing power or maximum RAM without the frills. Google decides the hourly pricing based on the number of virtual CPUs and RAM measured in the multiples of 1GB. Each vCPU per hour costs $0.03492 while 1GB of memory is $0.00468. Customers can also choose to launch the VMs in preemptible mode which is cheaper than the regular, on-demand pricing.

In 2013, I published an article at Gigaom Research on five features that AWS must fix. Here is an abstract from that post:

It is time for Amazon to switch to dynamic instance types, where customers are allowed to drag the sliders to select the memory, cores and CPU, and disk. This will simplify dealing with Amazon EC2 and put the customers in control of the server configuration. They can stop, tweak the configuration, and relaunch Amazon EC2 instances until the performance is satisfactory.

This was much before the general availability of Google Compute Engine. Two years later, Google added exactly the feature that I wanted to see in Amazon EC2 – sliders for configuring CPU cores and memory.

Even the hard core critics of Google agree that it has one of the best cloud platforms in the industry. When it comes to performance and value, Google Compute Engine is way ahead of EC2, Azure VMs, and SoftLayer. Features such as per-minute billing, shared storage, sustained usage discounts, and preemptible VMs are some of the examples of how Google’s innovating in the IaaS segment. Though IBM SoftLayer offers dynamic configuration of VMs, it doesn’t have the advantage of preemptible VMs and sustained usage discounts.

In its latest effort to outsmart competition, Google pulled of a coup with GCE custom machine types feature that enables customer to mix and match compute resources to design the most optimal configuration. The true value of cloud is realized when customers can scale their workloads horizontally and vertically. Custom machine types put the control in the hands of customers by letting them choose what they need. This move is certainly a game changer in the IaaS market.