https://www.techtarget.com/whatis/feature/Breaking-down-the-cost-of-cloud-computing
A large number of companies have migrated to a cloud environment in recent years, chiefly in search of scalability and cost savings. Cloud computing promises businesses certain financial advantages, such as avoiding on-premises facilities and equipment maintenance. Yet more than a few enterprises suffer sticker shock from unexpected charges, so it's important to understand cloud computing's different uses and how usage affects pricing and, in turn, a company's bottom line.
Everything used in public cloud is metered. Cloud service providers (CSPs) take into consideration several factors when determining how much to charge customers, including the following:
The big three CSPs -- Amazon Web Services (AWS), Microsoft Azure and Google Cloud Platform -- offer multiple layers of products and services to support user workloads depending on the user's needs. After extensive research into team collaboration products, we present a sampling of three leading cloud-based services that have a variety of storage and compute options along with examples of their pricing.
AWS offers businesses a pay-as-you-go (PAYG) pricing option for most of its cloud services. By paying only for the services used, companies avoid long-term contracts or complex licensing without additional costs or termination fees.
Amazon EBS is a block storage service for Amazon Elastic Compute Cloud (EC2). Businesses pay only for what they use until they release the storage.
| Volume type | Price per provisioned storage |
|
Free Tier: 12 months free |
30 GB of storage, 2 million I/Os and 1 GB of snapshot storage |
|
General Purpose SSD (gp3): Storage |
$0.08/GB-month |
|
General Purpose SSD (gp3): IOPS |
3,000 IOPS free and $0.005/provisioned IOPS-month over 3,000 IOPS |
|
General Purpose SSD (gp3): Throughput |
125 MBps free and $0.040/provisioned MBps-month over 125 MBps |
|
General Purpose SSD (gp2): Volumes |
$0.10/GB-month of provisioned storage |
|
Provisioned IOPS SSD (io2): Storage |
$0.125/GB-month |
|
Provisioned IOPS SSD (io2): IOPS |
$0.065/provisioned IOPS-month up to 32,000 IOPS |
|
Provisioned IOPS SSD (io1): Volumes |
$0.125/GB-month of provisioned storage and $0.065 per provisioned IOPS-month |
|
Throughput Optimized HDD (st1): Volumes |
$0.045/GB-month of provisioned storage |
|
Cold HDD (sc1): Volumes |
$0.015/GB-month of provisioned storage |
Amazon EFS is a serverless, elastic storage service for Amazon EC2. With EFS, organizations don't pay a minimum fee or setup charge. Companies only pay for the following:
The following table shows storage pricing for file systems using Amazon EFS lifecycle management for select regions. There's also a free-tier option available for one year.
| Region | Effective storage price ($/GB-Mo)-One Zone* | Effective storage price ($/GB-Mo)-Standard** |
| U.S. East (N. Virginia) | $0.043 | $0.08 |
| U.S. West (Oregon) | $0.043 | $0.08 |
| Africa (Cape Town) | $0.054 | $0.10 |
| Asia Pacific (Seoul) | $0.047 | $0.09 |
| Canada (Central) | $0.047 | $0.09 |
| Europe (Ireland) | $0.046 | $0.09 |
| AWS GovCloud (U.S.-East) | $0.056 | $0.11 |
Amazon S3 is an object storage service that stores data as objects within buckets. Users only pay for what they use, and there's no minimum charge. S3's Free Tier includes 5 GB of Amazon S3 standard storage.
Amazon S3 has five cost elements: storage pricing, data transfer and transfer acceleration pricing; data management and analytics pricing; request and data retrieval pricing; replication pricing; and the price to process data with S3 object Lambda. S3 storage classes and pricing vary by region. The pricing reflects the U.S. East (N. Virginia) region, and restrictions apply.
Amazon EC2 is a cloud compute service that allows users to spin up VM instances with the amount of computing resources they need, such as the number and type of CPUs, local storage and memory.
Users can try EC2 for free and receive 750 hours of Windows and Linux t2.micro instances every month for a year.
There are numerous ways to pay for EC2 instances:
AWS Lambda is a serverless compute service that enables users to run code without the need to provision or manage servers. Users only pay for what they use.
AWS Lambda pricing is calculated based on the following:
Azure offers several storage services, each with its own pricing model. Additional charges apply for snapshots, transactions on data and data transfer volume.
Example prices in the following table are for the East U.S. 2 region with locally redundant storage and are subject to change. For details on free services or additional charges for snapshots or data transactions, visit the Azure Blob Storage or Azure Files pricing sites.
| Data storage | Premium | Hot | Cool | Cold | Archive |
| First 50 TB/month | $0.15 per GB | $0.018 per GB | $0.01 per GB | $0.0036 per GB | $0.00099 per GB |
| Next 450 TB/month | $0.15 per GB | $0.0173 per GB | $0.01 per GB | $0.0036 per GB | $0.00099 per GB |
| More than 500 TB/month | $0.15 per GB | $0.0166 per GB | $0.01 per GB | $0.0036 per GB | $0.00099 per GB |
| Data storage | Premium | Transaction optimized | Hot | Cool |
| Data at-rest (GiB/month) | $0.16 per provisioned GiB | $0.06 per used GiB | $0.0255 per used GiB | $0.015 per used GiB |
| Snapshots (GiB/month) | $0.136 per used GiB | $0.06 per used GiB | $0.0255 per used GiB | $0.015 per used GiB |
| Metadata at-rest (GiB/month) | Included | Included | $0.027 | $0.027 |
Users pay for Azure VMs in four ways: PAYG, Azure savings plan for compute, spot instances and reserved instances.
Microsoft offers various types and sizes of VMs, divided into six categories: general purpose, compute optimized, memory optimized, storage optimized, GPU and high-performance compute.
Following is an example of pricing for compute- and memory-optimized Linux VMs in the East U.S. region.
Pricing for Google Cloud storage services is based on the volume of data stored, how long the data is stored, the number of operations the customer performs on the data, and any network resources the customer uses to move or access the data.
The following data storage rates apply to much of the United States:
| Standard storage | Nearline storage | Coldline storage | Archive storage |
| Starts at $0.020 per GB, per month | Starts at $0.010 per GB, per month | Starts at $0.004 per GB, per month | Starts at $0.0012 per GB, per month |
In addition, users are charged if they retrieve cold storage data or delete cold storage data early. That's because cold storage classes are for storing long-term, infrequently accessed data.
The following resources are available to users of the Google Cloud Free Tier, but specific limits apply:
| Resource | Monthly free usage limits* |
| Standard storage | 5 GB-month |
| Class A operations | 5,000 |
| Class B operations | 50,000 |
| Data transfer | 100 GB from North America to each GCP data transfer destination, Australia and China excluded |
In Google Compute Engine, VMs are grouped according to types of workloads, including general purpose, accelerator optimized, compute optimized and memory optimized.
Following are prices for compute- and memory-optimized machine types:
Compute-optimized machine types. Pricing per vCPU and gigabyte of memory for C2 machine types in Iowa (U.S.-Central) is as follows:
| Item | On-demand price | 1-year resource-based commitment price | 3-year resource-based commitment price | 1-year flexible CUD consumption rate | 3-year commitment |
| Compute-optimized vCPUs | $0.03398/vCPU hour | $0.02141/vCPU hour | $0.01359/vCPU hour | $0.24467/vCPU hour | $0.0183503/vCPU hour |
| Compute-optimized memory | $0.00455/GB hour | $0.00287/GB hour | $0.00182/GB hour | $0.00328/GB hour | $0.00246/GB hour |
Google also offers H3 compute machines. H3 VMs are powered by the 4th generation Intel Xeon Scalable Processors, DDR5 memory and Google's custom Intel Infrastructure Processing Engine. Pricing per vCPU and gigabyte of memory for H3 machine types in Iowa (U.S.-Central) is as follows:
| Item | On-demand price | Spot price* | 1-year commitment | 3-year commitment |
| Compute-optimized Cores | $0. 04411 / Cores hour | Unavailable | $0.03617 / Cores hour | $0.022647 / Cores hour |
| Compute-optimized Memory | $.00296/GB hour | Unavailable | $0.00243 / GB hour | $0.001789/ GB hour |
Memory-optimized machine types. The following lists pricing in Iowa (U.S.-Central) per vCPU and gigabyte of memory for M3 machine types, offering 3rd generation Intel Scalable Processors. Currently, M3 machine types are only available in certain regions and zones:
| Item | On-demand price | Spot price* | 1-year commitment | 3-year commitment |
| Memory-optimized vCPUs | $0.041064 / vCPU hour | $0.086494 / vCPU hour | $0.0224249 / vCPU hour | $0.012331 / vCPU hour |
| Memory-optimized Memory | $0.006018 / GB hour | $0.0012744 / GB hour | $0.0035636 / GB hour | $0.0018054 / GB hour |
There are three types of costs to consider when deploying on-premises infrastructure:
Building out and maintaining infrastructure on-premises means adding staff, which increases costs.
Typically, for every dollar a company spends on capital expenses to upgrade its IT infrastructure, there's an additional $2 to manage, maintain and secure that infrastructure.
To accurately calculate the total cost related to moving to the cloud, a company must conduct a thorough audit of the status of its IT infrastructure, including all direct and indirect costs. Direct costs -- software, hardware, maintenance, staff and the physical facility -- are easier to calculate. Indirect costs like the loss of productivity for reasons ranging from server downtime to addressing customer displeasure or repairing a damaged reputation are more difficult to determine.
The company then calculates the estimated cloud infrastructure costs using various pricing calculators, such as the in-depth monthly cost calculator provided by AWS.
Migration to the cloud historically has come with many benefits, including fast implementation, scalability, reduced costs, quick data recovery and fewer wasted resources:
In addition to the administrative benefits, moving to the cloud offers considerable financial incentives, including the following:
Public cloud providers often offer free or low-cost signups. Sometimes, cloud vendors even help their customers move their data to the cloud for free.
Money might not be a factor, but time spent is. It can take weeks or even months to transfer petabytes of data to the cloud. During that time, a company could be unable to access its critical data.
Transferring data from one cloud vendor to another can be troublesome and expensive, which is why most businesses stay with one provider. Vendor lock-in results because of the difficulty in canceling existing provider relationships.
In addition, many public cloud vendors charge a nominal fee every time a customer accesses its data. Businesses sometimes fail to account for these small fees when they consider the ROI of cloud computing.
Furthermore, data use increases translate into business cost increases for companies running thousands of analytics jobs. As such, using the public cloud for everything is a poor long-term strategy.
Still, successful businesses recognize the following aspects in making cloud computing decisions:
Businesses considering a move to the cloud must do a needs assessment to plan the migration and gain buy-in from stakeholders. After evaluating its current IT infrastructure, the IT team engages in a structured discovery process with stakeholders through meeting and to ensure everyone identifies their business needs when the IT team selects the proper cloud services and cloud providers. The IT team reviews the proposed cloud services to pinpoint where business requirements match.
After determining the appropriate cloud service and service provider, the IT team presents this information to the business leaders, detailing the proposed options and the reasons for selecting those options. These stakeholders can then decide whether to keep the operation on-premises, move it to the cloud or embrace a hybrid cloud approach, which uses a mix of on-premises and cloud services.
Andy Patrizio is a technology journalist with more than 30 years' experience covering Silicon Valley who has worked for a variety of publications -- on staff or as a freelancer -- including Network World, InfoWorld, Business Insider, Ars Technica and InformationWeek.
13 Dec 2024