Skip to content. | Skip to navigation

Personal tools
Log in

DIstributed Computing Environments (DICE) Team
We are a group of computer scientists and IT experts from the Department of Computer Science AGH and ACC Cyfronet AGH. We are a curiosity- and research-driven team, specializing in large-scale distributed computing, HPC, Web and Cloud technologies. We develop new methods, tools and environments and we apply these solutions in e-Science, healthcare and industrial domains.


Sections
You are here: Home DICE Blog Can cloud computing be free?

Can cloud computing be free?

Posted by Maciej Malawski at Jan 31, 2012 12:00 PM |
Everybody knows that one of the features of cloud computing is "pay-per-use". People are convinced, that although prices may drop, cloud will never be free (see e.g. http://queue.acm.org/detail.cfm?id=1772130). On the other hand, Google, Amazon, Apple, Microsoft and many others are offering some cloud services for free. The question is how much we can actually get for free. This is what we tried to measure.

In a recently accepted paper for IEEE Internet Computing we evaluated whether it is possible to use free tier of Google App Engine for compute intensive applications. App Engine is a platform-as-a-service (PaaS) cloud mainly designed to deploy Web applications written in Python or Java. It automatically scales the number of running instances (worker nodes) based on the service load. However, it imposes some constraints on the application, e.g. not all system calls are allowed on the platform, as well as the processing time of a request is limited to 10 minutes (recently extended from its original 30 second limit). We found that using a task queue API of App Engine it is possible to deploy an application that processes a number of fine-grained tasks, such as Monte Carlo integration, and use the platform auto-scaling feature to achieve a considerable speedup over a sequential processing. Moreover, if we enable billing, the free qoutas are increased, so even a better speedup can be achieved for free. Of course, our experiments with free computing on Google App Engine are rather toy examples in terms of scale, but we can conclude that such a platform-as-a-service solution would be very convenient for scientific applications.

Amazon EC2 also offers a free acces tier for the first year. 750 hours a month allow to use continuously the t1.micro instance for free for one year. We found that the micro instance does not give much CPU power: our tests suggest that the VM gets only 5% share of CPU time. Possibly, launching 20 micro VMs in parallel could help achieving a performance of a single core.

Windows Azure is another cloud platform which offers free access for 90 days. We have not tested this cloud due to the fact that it would require more effort to port our Java-based application, but we can try to compare their free offerings based on their specifications. Free offer each month includes 750 CPU hours of Extra Small Compute Instance (1GHz CPU) and 25 hours of Small instance (1.6 GHz). This allows uninterrupted (100% of the time) usage of Extra Small instance and 3% of CPU time of Small instance. These specifications suggest that the performance may be similar or higher to the one offered by free App Engine (1GHz corresponds to 37% share of typical 2.66GHz server CPU) but the actual performance may vary.

But not only compute cycles come for free. Google App Engine offers 5GB of free storage, EC2 gives 5GB of free S3 space and 30 GB of EBS storage with the free tier. Microsoft also gives 20GB of storage and 1 GB for a database. All these options are best suited for use by applications, but there are simpler solutions as well. For example Dropbox.com gives 5GB of storage that is automatically synchronized between the machines (laptop, PC, tablet) and allows easy file sharing.

We may wonder if these resources are enough to do science. Probably not, if you are a high energy physicist and need to process the data from LHC Atlas detector. But if you are a bioinformatics PhD student and need to do some advanced but not so computationally demanding simulation or data analysis tasks, such cloud resources may be just for you. And they may be simpler to use than the servers your department provides.

 

References

Malawski, M.; Kuzniar, M.; Wojcik, P.; Bubak, M.; , "How to Use Google App Engine for Free Computing," Internet Computing, IEEE (accepted), doi: 10.1109/MIC.2011.143
URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6065729&isnumber=5226613

Comments (0)

NOTE! This web site uses cookies and similar technologies (if you do not change the browser settings, you agree to this).

cyfronet agh