WHO MOVED MY CLOUD? Part II: What s behind the cloud vendors AWS, Azure and GCP? Part II of an ebook series covering cloud infrastructure and platform fundamentals not to be missed when preparing to make an organizational move to the cloud. In this part, AWS vs. Azure vs. GCP are covered for DevOps and IT Managers. Introduction Cloud vendor competition is heating up to include a pricing war and new feature offerings in the battle for your business. To better assist in the decision making process, this ebook offers a concise breakdown of the three market leaders: Amazon Web Services (AWS), Google Compute Platform (GCP) and Windows Azure (Azure), who, according to Gartner, together hold the majority of the IaaS market in 2015. While AWS is the clear leader in the market (as it has been since its inception in 2006), Azure is the fastest growing cloud provider, with triple digit growth in 2014 and 2015. GCP, far behind in market share, is still considered a top visionary by Gartner based on the completeness of their offering, go-to-market strategy, enhanced performance and global infrastructure. What's In a Cloud Vendor? Despite Amazon s significant head start, other players have entered the market, making their mark, moving up the ladder and influencing pricing, features and expectations. Azure has been catching up with their cloud solution, Google entered in late 2013 with its Google Cloud Platform offering, along with many other players, providing broad or niche solutions for different businesses. www.cloudyn.com @cloudyn_buzz cloudyn.com/blog August 2015
AWS vs. Azure vs. GCP: Features Instance Types AWS compute service, the Elastic Compute Cloud (EC2) offers the most diverse offering splitting its instances into nine families with 38 instance types (with an additional 15 older generation instance types still being offered). The families offer burstable, general purpose, CPU-optimized, RAM-optimized, GPU-optimized, storage-optimized, and more. Azure provides six families with 25 instance types, starting with a basic tier, which offers baseline performance for single instances (no scalability) and a standard tier, which offers general purpose, CPU-optimized, RAMoptimized, and network-optimized. GCP has the most focused offering, splitting its instances into four families (burstable, standard, high-memory and high-cpu) with a total of 18 instance types. Block Storage AWS block storage service is called Elastic Block Storage (EBS) and supports three types of persistent disks: Magnetic, SSD and SSD with provisioned IOPS. Maximum volume sizes range from 1TB for magnetic disks, up to 16TB for SSD disks. Azure s block storage service is called Page Blobs and supports standard storage (magnetic) and premium storage (SSD), with up to 1TB per volume. GCP offers two types of persistent block storage: magnetic and SSD. Volumes can be as large as 10TB, though Google recommends sizing the volume according to required performance (which drops as the volume grows). Object Storage The most popular object storage is AWS Simple Storage Service (S3), with storage options offering three different levels of redundancy: standard, reduced redundancy and Glacier (for archiving). Azure s object storage service is called Block Blobs (not to be confused with block storage) and offers local-, zone- and geographic-redundant storage. GCP offers standard, durable reduced availability, and nearline (for archiving, with very low access time). 2
I/O Speed When configuring a deployment, there are multiple parameters that can affect speed. These include the number and generation of CPU cores, number of instances, network speed, caching, and storage type. While speed is important, it is only one parameter out of many that should be considered in vendor selection, so benchmark test results should be taken with a grain of salt. Overall, based on a test conducted by InfoWorld in 2014 (image below), of the three vendors, GCP is fastest, followed by AWS and Azure. Global Network AWS has the widest range of region availability with eleven regions: Three in the US, two in the EU, two in East Asia, one in China, one in Australia and one in South America, as well as another region in the US for the sole use of US government agencies (AWS GovCloud). Each of these regions are comprised of multiple data centers or Availability Zones and edge locations. They use private networks for the data centers connectivity within a region, and the Internet for inter-region connectivity. 3
Azure runs 17 data centers, referred to by Microsoft as regions across Asia, Australia, Europe and North America. Lacking their own network infrastructure, they use the Internet for data center connectivity, labeling the data with QoS tags. GCP has only three regions: Central US, Western Europe, and East Asia. Each of these regions are comprised of multiple data centers or Zones. Contrary to AWS and Azure, however, connectivity between data centers is done based on Google s private global network, both on the regional level, and between regions. 4
AWS vs. Azure vs. GCP: Pricing Models Calculating instance cost is a complex task which requires looking at compute power, memory resources, and networking needs, then attempting to calculate project lifecycle needs for proper capacity, resource and personnel planning. The various tools provided by the three vendors aim to make this process easier, but it still leaves many in the dark. Billing Granularity AWS charges cloud usage on an hourly basis, meaning that even if a full hour of usage wasn t completed, the price will be rounded up. GCP and Azure both charge based on one minute intervals, with Google having a ten minute minimum to start. AWS: On-Demand, Reserved and Spot Instances AWS instance types and prices break into three categories: On-Demand, Reserved Instances (RI) and Spot Instances. On Demand The most common instance pricing model used, and the most expensive, with the highest cost per hour charged for usage out of the three pricing models. Full price is paid for utilization and doesn t include persistent storage and networking costs. With no long-term commitments or upfront fees, it s the most flexible option which may be ordered on the fly when needed. Reserved Instances A reserved instance is an upfront commitment made by the user for usage of a specific instance type, with a specific OS, operating from a specific Availability Zone. Commitment is made upfront for one or three years, and payment is made up front in monthly installments or some combination of the two. In return for the user s commitment, Amazon guarantees the reserved capacity at any time during the RI s life, as well as significant discounts over On-Demand rates. The three types of RIs are: All-upfront payment for the whole duration of the RI is made on day zero, with no monthly installments. This carries the highest discount rate. 5
Partial-upfront partial payment (about two-thirds) is made up front and the rest is divided into monthly installments over the RI s lifetime. This option carries slightly (1-2%) lower discount rates compared to all-upfront. No-upfront zero upfront payment, monthly installments only. This option is the least cost effective, with 10-11% lower discount rates than all-upfront. The Reserved Instance pricing model deviates somewhat from the pay-as-you-go model so synonymous to cloud services, and some companies view it as a CapEx investment as opposed to OpEx, which undermines original motive to migrate to the cloud. Before taking the Reserved Instance path, one must estimate the instance s planned utilization and calculate a break-even point of utilization, after which the RI investment yields a positive ROI: Reserved Instances compared to On-Demand pricing. All- and Partial-upfront RIs are almost indistinguishable and indeed represent very small rate differences Spot Instances A spot instance is acquired through bidding in an auction with Amazon determining the price based on supply and demand. This is the least expensive pricing model for EC2 instances (discount around 90%) however the instance can be terminated by AWS at any time, with just a two-minute warning before termination. 6
Microsoft Azure: On-Demand and Enterprise Agreements On-Demand Same as the AWS On-Demand pricing model. Enterprise Agreements (EA) Microsoft customers under Enterprise Agreements can use them not only for purchasing Microsoft software licenses, but also for purchasing Azure resources. Significant discounts are given over on-demand pricing in return for an upfront monetary commitment by the enterprise. Discounts vary from one Microsoft customer to another, so be sure to check your EA for precise pricing and rates. GCP: Sustained Usage and Inferred Instances On-Demand Same as the AWS and Azure on-demand pricing models. Sustained Use Sustained usage is a retroactive discount received at the end of the month for using an instance for an extended period of time during the month. It s basically a discount which is awarded without any upfront commitment or payment. Sustained Use discounts are tiered as follows: 7
Utilization Tier Pricing 0-25% of the month 100% of on-demand rate 25-50% of the month 80% of on-demand rate 50-75% of the month 60% of on-demand rate 75-100% of the month 40% of on-demand rate In example, an instance running 100% of the month will be charged an average of 0.25 100%+0.25 80%+0.25 60%+0.25 40% 0.25+0.25+0.25+0.25 = 70% of the base on-demand rate (or 30% discount), whereas an instance running 65% of the month will be charged an average of 0.25 100%+0.25 80%+0.15 60% 0.25+0.25+0.15 discount). =83% of the base on-demand rate (or 17% Inferred Instances When computing sustained use discounts, Google will give you the maximum discount available using Inferred Instances. An Inferred Instance combines multiple, non-overlapping instances of the same machine type in the same zone into a single instance for billing. With inferred usage, you are more likely to qualify for sustained use discounts. The figure below gives an example of grouping instances into Inferred Instances. 8
AWS RI vs. GCP Sustained Use Unlike AWS s RI instance, Google s Sustained Use instance does not require any upfront costs as the discount is applied at the end of the month retroactively. Additionally, businesses are not committed to use a specific instance for any specific amount of time. This does however make it more difficult to forecast a project s cost, as it is unclear until the end of the month. 9
Summary Table AWS Azure GCP Features Instances: Families and Types 9 families; 38 types 6 families; 25 types 4 families; 18 types I/O Speeds Second place Third place First place Max. Volume Size Global Network Connection 1 TB 1 TB Up to 10 TB Wide regional offering, most data centers and access points Intra-region: Private network Inter-region: Open Internet Global data center infrastructure Open Internet Narrower regions, centers and access faster per connection Global fiber network Billing Pricing Granularity Models Per Hour rounded up On-Demand, Reserved, Spot Per minute rounded up On-Demand, EA commitments Per minute rounded up (minimum 10 minutes) On-Demand, sustained use discounts 10
Conclusion Companies preparing for cloud migration must do their homework to find the best route for them. Each vendor has its own sweet spot for particular deployments and it is up to the migrating company to figure out which KPIs are most important to them and select a cloud vendor accordingly. Which one will fit and integrate with business needs best, as well as how it meets business short term and long term goals. ABOUT CLOUDYN Founded in 2011, Cloudyn is the leader in cloud monitoring and optimization. The company s industry award-winning SaaS solution delivers unprecedented insights into usage, performance, and cost, coupled with custom prescriptive actions for enhancing performance and reducing cloud spend.with more than 10,000,000 virtual instances monitored Cloudyn helps businesses select the right mix of cloud vendors, increase operational performance, reduce cloud costs to bring them under optimum control, and capitalize on customer choice. More than 2,400 customers use Cloudyn s technology worldwide including F500 industry leaders in aerospace, infrastructure, consumer online travel services, IT management consulting, and manufacturing. For more information, interested parties may visit www.cloudyn.com. 11