What may seem like a cost-effective option in the cloud can quickly turn surprisingly costly at the enterprise level. Here’s how: Cloud-only vendors focus on selling business owners their own cluster. That’s great, as you would be the master of your own destiny, outside the realm of IT. To do this, you only need to sign up for two compute units. The strategy is for it to be that easy. Preloaded already are industry-standard benchmarks. There’s no need to load the data, queries are already in place. The potential for analytics teams are all there.
Cloud-only data warehouses claim they have prices that meet customer needs. In this scenario, the price of about $3/hr/cu looks great, and businesses feel like they can be the masters of their own destinies, with a solution well within their budget. But, then some problems arise. The first is performance. The complexity and diversity of real-world workloads is a lot more challenging than anticipated. To address this, perhaps indexing or other tricks are considered. But, it turns out that there are no tricks to address this. The cloud-only vendor's answer is to “buy more”.
If, for example, it’s determined that “5X” will better address their needs, the only way to achieve that is by doubling the cluster size. So much for elastic performance. This is why cloud-only data warehouses costs can scale so exponentially. And, unfortunately, once a business is already committed to the solution, they may be apt to try to make it work. The next time the system needs to grow to address performance, the cost will double again. And so on.
The second problem is concurrency. Each compute unit, each cluster collectively can only do ~10 concurrent queries. That’s simply not enough. To handle this, cloud-only data warehouses don’t tell you to buy more. They cleverly market it as “auto-scale”. So, as queries increase and the systems auto-scale to address peaks and valleys in usage, costs can once again increase astronomically.
This is when hard choices need to be made. Does the business stick with this size and a price tag they had never budgeted for, abandon having analytic capabilities, or seek other options?
Teradata has a few different ways to address this. Approaching the situation with a business—not IT—focus, Teradata can supplement the existing production system, the business can pursue an on-premises solution, or even a solution in the cloud, while migrating data throughout all. Best of all, Teradata offers Consumption Pricing. You only pay for what you use—period. Teradata meets customers where they are. It is 100% elastic.