The Cloud Report

Blog archive

Google's New Managed VMs Combine Best of IaaS and PaaS

Google is the latest major IT provider looking to gain wider usage of its enterprise cloud services with the addition of new Managed Virtual Machines, designed to offer the best attributes of IaaS and PaaS.

The company also added several new offerings to its cloud portfolio, and -- looking to make a larger dent in the market -- slashed the price of its compute, storage and other service offerings. 

Speaking at the Google Platform Live event in San Francisco, Urs Hölzle, the senior vice president at Google overseeing the company's datacenter and cloud services, talked up the appeal of PaaS services like Google App Engine, which provides a complete managed stack but lacks the flexibility of services like IaaS. The downside of IaaS, of course, is it requires extensive management.

Google's Managed VMs give enterprises the complete managed platform of PaaS, while allowing a customer to add VMs. Those using Google App Engine who need more control can add a VM to the company's PaaS service. This is not typically offered with PaaS offerings and often forces companies to use IaaS, even though they'd rather not, according to Hölzle.

"They are virtual machines that run on Compute Engine but they are managed on your behalf with all of the goodness of App Engine," Hölzle said, explaining Managed VMs. "This gives you a new way to think about building services. So you can start with an App Engine application and if you ever hit a point where there's a language you want to use or an open source package that you want to use that we don't support, with just a few configuration line changes you can take part of that application and replace it with an equivalent virtual machine. Now you have control."

Google also used the event to emphasize lower and more predictable pricing. Given that Amazon Web Services (AWS) and Microsoft frequently slash their prices, Google had little choice but to play the same game. That's especially the case given its success so far.

"Google isn't a leader in the cloud platform space today, despite a fairly early move in Platform as a Service with Google App Engine and a good first effort in Infrastructure as a Service with Google Compute Engine in 2013," wrote Forrester analyst James Staten in a blog post. "But its capabilities are legitimate, if not remarkable."

Hölzle said 4.75 million active applications now run on the Google Cloud Platform, while the Google App Engine PaaS sees 28 billion requests per day, with the data store processing 6.3 trillion transactions per month.

While Google launched one of the first PaaS offerings in 2008, it was one of the last major providers to add an IaaS and only made the Google Compute Engine generally available back in December. Meanwhile, just about every major provider is trying to catch up with AWS, both in terms of services offered and market share.

"Volume and capacity advantages are weak when competing against the likes of Microsoft, AWS and Salesforce," Staten noted. "So I'm not too excited about the price cuts announced today. But there is real pain around the management of the public cloud bill." 

To that point, Google announced Sustained-Use discounts. Rather than requiring customers to predict future usage when signing on for reserved instances for anywhere from one to three years, discounts of 30 percent of on-demand pricing kick in after usage exceeds 25 percent of a given month, Hölzle said.

"That means if you have a 24x7 workload that you use for an entire month like a database instance, in effect you get a 53 percent discount over today's prices. Even better, this discount applies to all instances of a certain type. So even if you have a virtual machine that you restart frequently, as long as that virtual machine in aggregate is used more than 25 percent of the month, you get that discount."

Overall, other pay-as-you-go services price cuts range from 30 to 50 percent dependency. Some of the reductions are as follows:

  • Compute Engine is reduced by 32 percent across all sizes, regions and classes.

  • App Engine pricing for instance-hours is reduced by 37.5 percent, dedicated memcache by 50 percent and data store writes by 33 percent. Other services, including SNI SSL and PageSpeed, are now available with all applications at no added cost.

  • Cloud Storage is now priced at a consistent 2.6 cents per gigabyte, approximately 68 percent lower.

  • Google BigQuery on-demand prices are reduced by 85 percent.

Google also extended the streaming capability of its BigQuery big data service, which initially was able to pull in 1,000 rows per second when launched in December to 100,000 now. "What that means is you can take massive amounts of data that you generate and as fast as you can generate them and send it to BigQuery," Hölzle said. "You can start analyzing it and drawing business conclusions from it without setting up data warehouses, without building sharding, without doing ETL, without doing copying."

The company expanded its menu of server operating system instances available with the Google Compute Engine IaaS with the addition of SUSE Linux, Red Hat Enterprise Linux and Windows Server.

"If you're an enterprise and you have workloads that depend upon Windows, those are now open for business on the Google Cloud Platform," said Greg DeMichillie, a director of product management at Google. "Our customers tell us they want Windows, so of course we are supporting Windows." Back in December when Google's IaaS launched, I noted Windows Server wasn't an available option.

Now it is. There is a caveat, though: The company, at least for now, is only offering Windows Server 2008 R2 Datacenter Edition, noting it's still the most widely deployed version of Microsoft's server OS. There was no mention if and when Google will add newer versions of Windows Server. The "limited preview" is available now.

Posted by Jeffrey Schwartz on 03/26/2014 at 9:30 AM


Subscribe on YouTube