Google had a Cloud Platform Live event on Tuesday (recorded here) unveiling a number of new features, improvements and a new pricing model.
Pricing
Considering that cloud computing prices should follow Moore’s law applied to hardware, and since the prices for hardware have had a 20 to 30% drop per year for the last several years, Google decided to make major cuts for several of their cloud services:
| Service | Price Cut by % | Comment |
| Compute Engine | 32% | across all regions, sizes and classes |
| Storage | 68% | current price 2.6¢/GB |
| BigQuery | 85% | $5/TB for on-demand usage |
Google has introduced a Sustained Use Discount model, lowering prices in steps for using instances more than 25% of a month’s time, resulting in a 30% cut for those instances running the entire month. For such an instance, the total price cut is 53% when compared to the old pricing model. The advantage of the Sustained Use Discount model is that besides offering variable price cuts based on usage, users do not have to decide upfront how much instance they are going to use over the month. The price is automatically cut as one progresses throughout the month based on how much the instance runs, 25%, 50%, 75% or 100% of the time.
RightScale, a service provider across multiple cloud platforms, has performed a detailed comparison of Google’s new prices against Amazon’s, showing that GAE is from 37.78% to 60% cheaper than AWS for on-demand instances. For sustained use (GAE) vs. 1-year reserved instances (AWS), RightScale noticed that Google is cheaper by 21.18% to 32.05%, depending on the instance type. For 3-year AWS reserved instances, Google’s prices are higher than Amazon’s, somewhere between 3 and 18%, but RightScale observed:
It is important to note that buying AWS reserved instances requires a one-year or three-year commitment for the same instance type with the same operating system in the same region. If AWS hourly on-demand rates go down during the three years, you will not get the lower price. AWS does sometimes reduce hourly pricing for reserved instances, however, looking at historical precedent, AWS does not always apply these savings to your already purchased RIs. In contrast, Google sustained-use pricing is calculated as a percentage of the on-demand baseline rate. As Google’s baseline rates go down, the sustained-use prices will fall as well.
As a result, all future Google price drops are passed on to all customers immediately when they take effect.
Developer Tools
Google App Engine now includes integrated tools for building, testing and releasing apps into the cloud. For code repositories, it supports git/GAE or GitHub, while for CI it includes support for Maven, Gradle, and Jenkins. They demonstrated during the live session how one can commit a fix to a main repository and see the code going through automated building, testing and deploying.
They also added support for live migrations demoing how an instance running a video was migrated to a different geographic location in the cloud, with a preparation time of about 20-30 secs while the actual switch between the old and the new instance, the only part noticeable by the user, took ~1 second. The new instance has the same IP and configuration, everything is the same as for the old one, the only difference being a log note mentioning the switch. Google did not detail how that was made possible, but a video app can switch easily having to just start streaming at a specific time in the video. For other types of applications, switching might be a lot more complex. Nonetheless, the live migration was impressive and done quite swiftly.
GAE now has aggregated logs across all instances along with filtering and search, and detailed stack traces for debugging purposes.
Operating Systems
Google Compute Engine has added support for Windows Server 2008 R2, SUSE Linux Enterprise Server and Red Hat Enterprise Linux.
Google also announced an integrated Cloud DNS service.
Managed VMs
Currently in Early Access, Managed VMs provide the benefits of Compute Engine flexibility and GAE’s scalability and auto-management. A managed VM is desirable when when a resource is not available in GAE or when one wants to access native code or resources. After the VM is built, it is then integrated into App Engine, having access to various services such as MemCache, Datastore, Task Queues, and it is automatically managed.
BigQuery
BigQuery now supports Streaming using tabledata().insertAll() at the rate of 100K records/sec/table, enabling large data stream analysis in almost real time. The pricing is 2.6¢/GB/month for static data analysis (data uploading, exporting and table reads being free), and 1¢/100K-records for streaming. For on-demand querying Google charges 0.5¢/GB, both for interactive and batch queries. For large workloads, reserved query capacity starts at $20,000/month, Google claiming that the price is 75% cheaper than those of other providers.
It remains to see how other cloud vendors, especially Amazon, will react to Google’s price cuts.
