28 January 2014 Posted by Paul Burns
Some interesting news from Cloudyn crossed the wire today and I had a chance to get some of the details from the company’s CEO, Sharon Wagner. Cloudyn, a company that provides services for managing and monitoring cloud costs for Amazon Web Services (AWS), has announced support for Google Cloud Platform (GCP), allowing customers to optimize their GCP spending the same way they do with AWS.
In addition to gaining support for GCP, customers can now compare instances between AWS and GCP. The updated offering from Cloudyn gives recommendations on fitting workloads to the right instance size and type within AWS (for Elastic Compute Cloud (EC2) instances) and GCP (for Google Compute Engine (GCE) instances). It also recommends porting workloads between the two clouds based on price, performance, and location requirements. The full details can be found in the company’s press release, here.
What This Means
While GCP lags behind AWS in its breadth of services – as all the other public cloud platforms currently do – GCP is widely considered a strong potential competitor to AWS. A few factors have played in to this perspective: Google certainly knows how to do big infrastructure and early feedback on the services is quite positive. They also have the bucks and appear willing to commit to infrastructure as a service (IaaS) and more. Time will tell.
Since general availability of GCP was announced in December 2013, prospective GCP customers have been asking questions like:
- Which cloud is cheaper, AWS or GCP?
- Which cloud is faster?
- Which cloud is more reliable?
- And of course, which cloud is best?
We don’t have answers to all these questions yet, but the ability to compare GCP and AWS instances using Cloudyn sheds some light on price and performance advantages and provides guidance on assigning diverse workloads and applications to the right cloud.
Context
Research by Neovise in 2013 showed that many cloud users choose clouds from multiple service providers. 33% of organizations that use IaaS reported that they had implemented 2 or more public clouds. One of the key reasons for this is that cloud customers want the best fit for their individual applications.
Keep in mind that there is no single best cloud platform for all your applications; cloud providers offer diverse features and benefits that may provide efficiency and cost savings for one workload, while hindering another. Users want the “best” cloud. Not the best cloud overall (which doesn’t exist), and not even the best cloud for “most” of their applications… They want the best cloud for individual applications on a case-by-case basis.
Some of the questions that cloud customers are asking mirror the questions that prospective customers have about GCE – for example, questions related to price, performance, price/performance ratios, and reliability. Cloudyn has also done some of its own research on the topic by surveying its large customer base. They found that customers consider price efficiency, performance and location to be some of the most influential factors impacting their purchase decisions.
Choosing the Best Cloud
The new GCP measurement capabilities from Cloudyn represent a step forward in terms of right-sizing and best-fitting workloads and instances. To validate its expanded services, Cloudyn offered 500 customers the chance to participate in beta testing. The result: A little more than half of these users found GCE more attractive for some of their workloads.
A portion of these results can be attributed to the fact that GCE instances deliver better ROI for I/O intensive workloads than EC2 instances. GCE instances are built on all SSD components and have much faster I/O by default. While AWS offers high-IOPS instances that match this performance, they cost much more.
But AWS isn’t without its own benefits. During its AWS re:Invent conference in November last year, Amazon announced its new compute-optimized instance family, C3. This family of instances may be the best price/performance value in AWS and the comparable Google instances tend to be more expensive.
Cost and performance comparisons can also get significantly more complex. Since GCP does not currently offer reserved GCE instances, reserved EC2 instances from AWS can yield cost advantages in some cases. Yet, Google has price and performance advantages for high IO workloads. This means that the full costs, in terms of both price AND performance must be netted out against each other to find the best cloud and the best instance size.
As a more complex example, Cloudyn had a case with one AWS customer running map-reduce jobs on m1.large instances. Typically about 800 instances were launched during the job, and the runtime duration for the instances was about 80 minutes. Cloudyn calculated that the same workload on Google would typically terminate in about 50 minutes instead of 80. So it looked like the compute cost on GCP would be less than half of that on AWS (factoring in shorter run times on GCP, lower instance costs and sub-hour billing). However, by using Spot Instances on AWS, which are heavily discounted, the total cost difference between Google and AWS was only 10% in favor of Google.
Looking Ahead
As many of you know, there are a number of other cost monitoring and management tools available today, such as those from Cloudability and Datapipe (recall that Datapipe acquired Newvem). We expect tools and services like these to increasingly focus on multi-cloud comparisons, with multi-cloud visibility and control becoming the norm over time.
Cloud brokerage solutions inherently encompass a multi-cloud perspective and we expect their multi-cloud capabilities to continue improving rapidly. For example, Gravitant, with its cloudMatrix software that enables IT organizations to become cloud service brokers, already provides multi-cloud comparisons for Amazon, Amazon GovCloud, Terremark, CenturyLink and GoGrid. The cloudMatrix solution does price comparisons and also makes recommendations for placing applications on the best fitting cloud based on governance requirements, SLAs, price and more.
With so much cloud comparison data becoming available, many new insights will emerge. For example, the pros and cons of sub-hour billing – notably advanced by cloud service provider ProfitBricks – should become fully understood as customers compare Google’s by-the-minute pricing with Amazon’s reserved instances.
By monitoring tens of thousands of AWS instances, Cloudyn has already determined that 65% of them run for less than two hours. While AWS customers end up paying for a full two hours of service, GCE and ProfitBricks customers only pay for the period of time that they actually use their instance. The cost differences add up quickly when, for example, 1,000 Hadoop instances are launched and retired within an hour or two.
There are plenty of other strong public cloud platforms out there, and some of them have important advantages over both GCP and AWS. We look forward to seeing more comparison data for these clouds. We also look forward to getting hard data that shows which clouds are the best match for each type of application.
]]>
Recent blog posts
- Revisiting the Benefits of SaaS
- Compliance Drives Buying Decisions in the Cloud
- Changing Competitive Dynamics for IaaS Providers
- Good News for Cloud Job Seekers in 2015
- Colocation: The New Center of Hybrid IT
- The Problem with Forced Cloud Reboots
- OnApp Federation Grows Stronger with SolusVM Acquisition
- HP to Acquire Eucalyptus: Questions, Comments and Concerns
- Is It Really ‘Game Over’ for Most IaaS Providers?
- The Big Cloud Winners Are…Not Just the Hyperscale Players