This article is more than 1 year old

Google's new cloud CRUSHES Amazon in RAM battle

Henry Ford 2.0 has a new competitor, and punters should rejoice

If Amazon is an arms dealer, then Google is an army that happens to also sell guns. For this reason, Google's general-availability release of its IaaS cloud is a key moment in the evolution of as-a-service IT, as it represents the arrival of a truly competitive market for rentable compute and storage.

With Monday's launch, Google has introduced new instances, cut prices, and broadened Linux support for its cloud. So far, so usual, but an analysis by El Reg indicates that this release poses a real competitive challenge to cloud kingpin Amazon.

Though Amazon operates a broader range of services and instances, Google's new rentable servers are a good proposition for thrifty developers, as they provide more RAM-per-dollar than Amazon, on average. Getting a decent RAM allocation is one of the best ways devs can speed up cloudy apps, such as databases, without having to invest in either dedicated hardware or high-end instances.

Google's competitiveness here reflects the architecture of the company's overall cloud, which favors a multi-million server scale-out approach that gives each node reasonable RAM, a few reliable processors, and blazing fast flash storage.

This compares with Amazon, which fields some of these capabilities but also sells a broader range of high-CPU, high-network, and other SKUs.


Google's new release puts further pressure on AWS's RAM prices

By offering developers a lower overall RAM cost per-instance, Google has made its rentable servers attractive to applications that need to thrash a low-latency, high-response memory tier.

Google has also stepped up to take on Amazon with high-CPU instances, which should force price competition in an area crucial for as-a-service supercomputing – a nascent technology that holds great promise for cash-strapped academics that can't get the necessary hardware from their university.


Amazon is still the leader for vCPUs

Just as Amazon has a few pet examples for cloud supercomputing, such as its regularly trotted out PR case study of cycle computing, so too does Google – but this update shows Google is also serious about offering these capabilities as a commercial endeavor rather than as an easy way to get headlines.

Amazon, though, is far more competitive than other providers in terms of how many virtual CPUs it can make available per instance. This, however, comes with the cost of the hugely unpredictable performance properties of AWS. Google is a more stable system, so although it fields fewer vCPUs, they may have more predictable performance.

The release also means that "the big three" clouds of Microsoft, Google, and Amazon are all now in general availability for core compute services. This will lead to a new phase of competition among these companies to win enterprise accounts, and should hopefully lead to further price cuts for the punter.


It takes a ton of resources to operate a competitive cloud

An analysis by El Reg earlier this year of the cumulative capital investment of these companies to date indicated that both Microsoft and Google have huge stores of as-yet-unutilized IT to draw on in their fight with Amazon, posing real competition for Bezos & Co., who have been unworried by smaller competitors such as Rackspace.

Perhaps most importantly, out of Microsoft, Amazon, and Google, the latter is by far the most technologically advanced firm. Amazon may have the experience, and Microsoft may have the knowledge that comes from supporting a range of diverse services, but Google employs the most capable people and operates some of the more advanced technologies. We already have evidence of this in flashy cloud tools such as live VM migration and SDN-enabled network reconfigurability, but given the mammoth capabilities of the upcoming "Omega" scheduler, we expect to see much invention out of the ad-slinger.

Google's launch stands in stark contrast to HP, whose own "general availability" cloud release brought with it more "known issues" than "features". HP's OpenStack-based cloud is not able to spin up more than 100 instances at a time without having problems, yet in a promotional video Google demonstrated the concurrent launch of 1,000 apparently healthy VMs. So far, El Reg has seen no evidence that OpenStack-based clouds can be more competitive either on price or performance than the proprietary systems fielded by the big three, and Google's launch stresses that point yet again.

Though cloud computing is a relatively niche part of the overall IT landscape, it is expected to gobble up an ever-larger share of IT spend. For this reason, El Reg's floating cloud bureau welcomes Google's cloud update, and looks forward to another year of brutal reprisal price cuts by its (increasingly-unnerved) rivals. ®


Reg readers: if you find this type of analysis useful, then let us know, and if there's other data you'd like us to delve into, please make your voices heard in the comments.

More about


Send us news

Other stories you might like