Google Announces Compute Engine - At Last An Alternative to Amazon AWS
Written by Alex Armstrong
Friday, 29 June 2012
Google's Compute Engine is the first credible alternative to Amazon's AWS. Until today the cloud was Amazon - now at last it's a two-horse race.
The cloud is an attractive idea - use someone else's hardware and data center and get the advantage of their huge investment and just pay for the time you use. The difficult bit is doing it so that users can set up complex systems remotely.
Until today AWS was the only serious general purpose cloud system. Basically you could upload any virtual machine image and plug together infrastructure in a fairly platform-independent way. The alternatives were either too small to count, or too specialized.
Microsoft's Azure was initially a vehicle to promote ASP.NET, IIS and other Microsoft technologies. Only now is it beginning to diversify to provide a general purpose cloud capable of running Linux-based VMs and build complex infrastructure.
Google also has an existing cloud system, but it essentially runs applications written in Python and while this is powerful it is hardly general purpose.
Now Google has announced the Compute Engine, which is a general infrastructure solution. Essentially it allows users access to Google's global data centers. Just as Amazon's AWS is based on the hardware its retail operation needs, so too Google's Compute Engine is based on the hardware it uses to provide search services. The spare and standby servers will be rented out to developers who will be allowed to upload Linux-based VMs - currently Ubuntu or CentOs.
The system's characteristics are:
Launch Linux VMs on-demand. 1, 2, 4 and 8 virtual core VMs are available with 3.75GB RAM per virtual core.
Store data on local disk, on Google's new persistent block device, or on its Internet-scale object store, Google Cloud Storage.
Connect VMs together using Google's high-performance network technology to form compute clusters and manage connectivity to the Internet with configurable firewalls.
Configure and control VMs via a scriptable command line tool or web UI. Or create a dynamic management system using Google's Compute Engine API.
The advantages that Google claim are scalability and cost. Given that Google needs huge computing resources, it can promise to support large computational requirements. It claims to have applications that use 10,000 cores on the same hardware. Given there is so much powerful hardware on offer, costs can be lower. The announcement includes the line:
"...allows Google Compute Engine to give you 50% more compute for your money than with other leading cloud providers"
And we know who "other leading cloud providers" are!
A full price table is available and from this, for example, a single core image with 3.75GB and 420Gbytes of storage is $0.145 per hour, which is a lot more power than the same rate would buy at AWS.
Currently the system is open as a limited beta. You can apply at the Compute Engine web site - it only needs email, phone and a short note on what you plan to do with the Compute Engine.
If nothing else having a real competitor should make Amazon work harder at making AWS more powerful and lower cost.
Algorithmia is a new marketplace for algorithms where people who need to use an algorithm can find one without having to implement it for themselves and algorithm creators can get paid for their effor [ ... ]
A video from SIGGRAPH 2014 presents a fully automatic approach to realtime facial tracking and animation which doesn't require calibration for different individuals and seems suitable for deployment i [ ... ]