Overview
It’s true that cloud is a strategic investment for any enterprise IT approach. However, there is no one-size-fits-all solution when it comes to analytics infrastructure design. As you evaluate cloud companies, it’s crucial to match a vendor’s offerings to your enterprise’s specific needs.
With so many cloud vendors offering a range of choices and pricing models, choosing the right cloud company for your needs can be challenging. Many cloud technologies are rapidly emerging onto the market, with expectations around “cloud-native” vendors particularly high.
It’s true that cloud is a strategic investment for any enterprise IT approach. However, there is no one-size-fits-all solution when it comes to analytics infrastructure design. As you evaluate the solution that works best for your enterprise’s needs, we recommend grading cloud companies against the following criteria:
1. Will They Actually Save You Money?
A key benefit that many cloud companies tout is low upfront cost and fast deployment times. But a cloud-native solution may not be cheaper for everyone, particularly for enterprise-scale workloads. Because cloud-only vendors scale hardware in the form of larger clusters or multiple clusters to solve problems like performance and concurrency, this process becomes inefficient and expensive at hyperscale levels.
Look for solutions with consumption pricing models, where you’ll have zero upfront commitment and the ability to pay as you go on a per-time-unit basis. You should only have to pay for the data storage and processing that you use. That means no long-term contracts that lock you in for months or years. No need to worry about utilization rates, capacity sizing or limits. You should also have usage transparency so you can allocate costs by department and set threshold alerts.
2. Do They Offer Flexible Deployment Choices?
In a survey of Teradata’s customers conducted in 2016, 90% of respondents said they wanted to deploy both on-premises and in the cloud by 2020. This prediction has proven to be accurate. Having the flexibility to run and migrate analytic workloads across different options provides tremendous operational agility. It also de-risks customers’ architectural decisions by allowing adjustment in how and where workloads are run as the business evolves. Flexibility and portability also grant customers freedom from lock-in because they can move anytime.
3. Do They Separate Compute and Storage?
You may encounter challenges if a cloud company offers a solution that scales storage and compute solutions simultaneously. Enterprises sometimes choose this method in order to manage massive amounts of data, thinking that they’ll maximize performance while reducing costs.
However, it’s important to be able to scale these capabilities independently. If data is streaming in at a rapid rate, having to scale compute nodes at the same level as storage could drive up costs when the value of the data is unknown. Or, demand might drive the need to scale up compute based on what the business user needs even as the amount of data being stored doesn’t change.
4. Do They Offer Flexible and Cost-effective Data Storage?
In recent years, organizations have chosen to store data on Hadoop hoping that this would be a cheaper and more flexible option. However, many enterprises are finding that Hadoop offers the complete opposite. It limits them to process data on-premises. And while Hadoop makes it easy to transfer data into an enterprise’s system, it’s difficult to get that data out when it’s needed.
Your vendor should provide native object storage allowing you to leverage all of your data, at lower cost and hassle. Look for object storage with unlimited scalability that is stored natively, which leads to easier management and better performance.
5. Do They Offer Enough Elastic Scaling Options?
Cloud companies tout the ability to scale elastically, but make sure that you have a wide range of scaling options. These can include the following:
- Scaling Up and Down – Change instance size, such as going from small to either medium or large instances, with just a restart without requiring data redistribution.
- Scaling Out and In – Adjust the quantity of compute instances without affecting storage and without requiring data redistribution.
- Stop/Start – Turn off compute instances to halt core operations for some period of time in order to optimize spend and then restart again when needed.
6. Do They Have a Core Analytical Production Capability?
Some companies may quickly design a cloud solution that they only use for testing and developing new products. But without a core analytical production capability, these companies lack the foundation they need to fully innovate. Look for core capabilities that provide end-to-end security and a foundation of proven and mature analytics production technology.
7. Do they integrate providers and tools effectively?
The enterprise must support a diversity of data types, tools, and even cloud platforms in order to keep up with the pace of innovation today. Harmoniously integrating various approaches will make your solution more powerful.
Look for an integrated solution that effectively brings together data lake, data warehouse, and analytics. Make sure that users across your enterprise and open source ecosystem can easily connect to the platform through APIs. They should be able to access a single trusted source of data, through a single seamless interface, and leverage the platform to run robust analytics that can supply downstream applications.
Your solution should also support multiple clouds. Native integration with cloud services enables the enterprise to run advanced analytics from a single environment that pulls in and joins data from other sources, such as Amazon S3 and Azure Blob.
The race is on to provide flexible, cost-effective, enterprise-scale analytics solutions. But you don’t need to limit yourself to cloud-native or cloud-only platforms — think more broadly about what your enterprise needs and aspires to achieve in the future. Then accept no compromises on your path to finding the right solution.