Starting nearly fifteen years ago, the advent of cloud computing infrastructure gave startups and departments in larger organizations a new way to build, test and deploy new software applications. It was a revelation.
This new public cloud amounted to a development nirvana that any software coder with a credit card or access to petty cash could use to rent the computer power, data storage, and network bandwidth needed for a project.
That model was a vast improvement over the traditional practice that typically required tens of thousands of dollars (often much more) to buy servers, storage, and software and months to years of development time even to update existing applications. This new public cloud was a dream come true especially for brand new apps.
Fast forward to today as more companies — giant corporations and government agencies in addition to small businesses — consider putting more workloads into the cloud, and unfortunately the reality doesn't always live up to their expectations. Here are some of the factors in modern cloud economics that company leadership should consider.
1 — Price out relevant cloud services
The three components that comprise most customers' cloud bills are compute nodes (CPUs), storage, and networking, so assess your needs across those services carefully. And, be aware that category price leaders of the past are not necessarily the best bargains today.
In networking, pay special attention to the cost of outbound data transfer if you need to move a lot of data from the cloud to customers, partners, or your own data centers. ‘Data egress' charges often lead to unpleasant surprises. It is almost always free to pump data into a given cloud — it's getting it out again that is expensive (and time consuming).
Also note that list price reductions for the most common cloud services have slowed to a crawl over the past four years. Many of these reductions have come to apply only to niche services or to geographic regions outside of North America and Europe. Price cuts to services you don't use in regions you don't serve will not cut your costs. The truth is that price leadership in the most popular cloud services has shifted in surprising ways.
2 — Know your data
While cloud is becoming the destination for more corporate data of all types, some companies may want to keep sensitive data inside their firewall for the time being due to regulatory requirements or other reasons. In that case, it will be helpful if your public cloud provider works well in tandem with your in-house tech. Ensuring smooth operations between internal and outside cloud services can save not just money but a lot of angst.
In addition, many companies want to run flexible pay-as-you-go cloud services in their own data centers, another reason there should be good fidelity between services running internally and those operating on the cloud provider's infrastructure.
In a perfect world, the two sets of infrastructure will mirror each other, but the reality is most public cloud providers offer only a subset of their services outside their own data centers. That could be a problem.
3 — Need for speed
In cloud, differences in cost often boil down to price/performance. If you're paying for both how much data you're processing as well as the time it takes to process transactions involving that data, it will cost you less to use a cloud that processes things faster. Put another way, the same amount of data processed will cost more on a cloud that has longer processing times.
When it comes to some applications, a slight delay between the time data is transmitted and when it is received is not a deal breaker. If email gets to its recipient in 15 seconds as opposed to 10 seconds, not many people will know or care.
But speedy execution counts in high-performance computing and online transaction processing scenarios, where data needs to be fed into and out of compute nodes fast. So, if you need super speedy data processing, make sure your cloud provider can provide that and oh, by the way, that will lower your cloud bill if your cloud provider charges by the drink because if your speeds are faster, your usage time declines, cutting your cloud bill
And, if your application requires high throughput —meaning that big chunks of data must be transferred and ingested at one time — that's another thing to vet with your provider before signing on.
In addition, if your company runs the key databases powering its manufacturing, inventory, and financial systems in-house, it may want to keep running affiliated applications in company-controlled data centers as well, easing compliance concerns and while also minimizing latency
4 — Location, location, location
As noted above, it takes time to transmit data even between two nodes in the same location. When you add significant distance to the mix, say sending data from Northern Virginia to Germany, latency rears its head again, even if the transit occurs between data centers run by the same provider. Keeping data close to where it will be processed or used saves time and money.
For this reason, make sure your cloud provider offers facilities in all the regions where you conduct significant business, making sure to check price differentials for the services you use as they can vary quite a bit by region.
In addition, global organizations should keep an eye on data sovereignty laws. These rules mandate that information about citizens of a given country remains within those borders, so make sure your provider runs cloud data centers in the countries where you operate, or offers a way to deploy public cloud services in the location of your choosing.
Consider your existing applications, the type of data you have, and both your performance and regulatory requirements before entering into a big cloud commitment.
Changing times require new thinking about resource allocation. For example, as demand for online video exploded due to the COVID-19 pandemic, cloud communications provider 8X8 found itself in need of more capacity to serve an influx of new video meeting customers — and lower costs because of the specific demands of its business. Vik Verma, CEO of 8X8, noted that Oracle Cloud Infrastructure was able to provide "the high level of security, the geographic reach, and the world-class support we needed to meet the astronomical surge in demand."
This pandemic will not be the last challenge to business-as-usual. Cloud computing comprises myriad services so look at those that are relevant to your business but make sure to scrutinize price on the key building blocks — computing, storage, and especially networking because that is typically where large unexpected expenses accrue.
And, while monthly cloud fees are a major component of overall price, slow performance of key applications, often due to latency, takes a toll in price performance. So paying for faster performance may end up costing less because you end up spending less time running pay-as-you-go resources.