When you’re buying cloud computing services from one of the big three – Google, Amazon or Microsoft – figuring out actual costs can be extremely complicated, since everything depends on your exact configuration and usage levels.
But savvy MSPs who are window shopping for cloud computing options on behalf of their clients should make sure they’re not missing out on the discount opportunities these providers offer, while considering the risk-reward nature of these programs.
Following are the most important factors to take into consideration for your customers:
Amazon Web Services (AWS)
For AWS, the discounts come as you reach certain tiers of service based on usage and payment method.
For example, storing more data means paying less per GB: Up to 50TB, it’s $.023 per GB monthly, reduced to $.022 per GB at 50TB-plus, before dropping to $.021 per GB at 500TB-plus.
That means a $600 annual reduction at 50TB and a $6,000 yearly drop at 500TB.
Most AWS discounts based on total usage, such as data transfers, kick in as milestones are reached and are automatically applied and calculated across all of the accounts on the consolidated bill.
So if you have two or three separate accounts with aggregate usage reaching a discount tier, you’ll realize those savings, which definitely means you should keep as many accounts as possible on the same bill.
AWS also provides a significant break when purchasing reserved instances vs. on-demand, which means you are committing to pay for (and potentially use) a certain amount of EC2 or RDS instances at the beginning of the year.
The benefit is saving money, but the risk is over-shopping for computing power for a year or three years.
For example, a one-year standard Reserved Instance term nets a 40 percent discount compared to on-demand EC2 pricing.
A three-year term increases that discount to 60 percent.
If you want to maintain the right to change instance families, operating system, tenancy or payment options, the three-year term RI discount is still available, but drops to 45 percent.
Layered on top of that is when you pay.
The more you pay upfront, the cheaper it gets.
For an EC2.m4large instance, you’ll save 32 percent by reserving it for the year but not paying anything upfront.
That leaves you owing $650 at the end of the year versus a $950-plus on-demand rate.
A partial upfront payment drops it to $552 per year – a 42 percent discount – while paying the whole thing upfront saves 43 percent, or $541 for the year.
Contracting for multi-year terms presents more discount opportunities, although your long-term commitment increases as well.
If you do jump on those discounts but don’t need those RDS instances after all, AWS does have a resale marketplace available to recoup some of your money.
Google Cloud Platform
One way to save with Google is qualifying for the “sustained use” discount.
If your instance is being used at least 25 to 50 percent of the month you’ll pay 80 percent, from 50-75 percent you’ll pay 60 percent, and if you pass 75 percent you’ll only pay 40 percent.
For an n1-standard-1 instance, that drops the hourly rate from $.0475 all the way down to $.0190 at the highest utilization level.
Much like AWS aggregating accounts under a consolidated bill, this discount also gets applied using inferred instances to maximize your savings across distinct instances, even if usage wasn’t actually concurrent.
Committed use discounts are another Google savings opportunity.
A one- or three-year commitment drops prices dramatically on vCPUs and memory.
For example, the $.033174 per hour vCPU price drops to $.011915 per hour with a one-year commitment, and to $.014225 per hour at three years.
Microsoft’s Azure offering has two fundamental purchasing and pricing methods: Pay-as-you-go and enterprise agreements.
Pay-as-you-go is a strictly usage-based payment model that charges for what you use and is available directly from Microsoft or its resellers.
Enterprise Agreements are where real discounts kick in, but as the name states, these are negotiated contracts with Microsoft or a reseller and subject to the whims of individual negotiations.
Due to reseller involvement and Microsoft’s enterprise sales models, specific discount details are harder to nail down and depend on myriad factors.
How do the discounts stack up?
For a Standard two CPU server with SSD memory, paying in advance saves money across the board: $832 for AWS versus $1,165 on-demand, $1,594 for Google versus $1,857 on-demand, and $699 for Azure, compared to $999 on-demand.
Memory likewise gets cheaper when paying upfront: AWS comes in at $104 per GB versus $149 on-demand; while Google charges $213 per GB versus $246 per GB on-demand; and Azure costs $100 per GB, compared to $140 on-demand.
Much like stocking up on fresh produce at Costco, you’d better be confident everyone’s going to be hungry, but paying in advance will save you money over the longer term if you’re fully utilizing your cloud processing power and storage.
However, the lock-in nature of these long-term discounts also means less flexibility if demand decreases and, for Azure, it may cost you significantly more if you need a mid-term increase.
Kirill Bensonoff is a seasoned entrepreneur and the founder of Unigma, a unified cloud management platform. Unigma has been featured in a number of publications, and Kirill blogs regularly about cloud, tech and growing your managed services business. He can be reached at [email protected].