The appeal of cloud computing infrastructure can be summarized in two words: Flexibility and Cost. The flexibility to spin up new hardware or use a new service on-demand when you need it. The cost benefit of paying only for what you consume, mitigating the need to purchase additional hardware for occasional surges in demand.
To really maximize the benefits of the cloud, midsize and large enterprises should look for applications that are built to take advantage of hybrid architectures. Hybrid means you can move your applications (in whole or in part) across public cloud providers and on-prem data centers without friction.
Let’s look at why this is important.
Flexibility in the context of IT means the ability to use new hardware, software and services when you need them and where you need them, with minimal friction. I will, however, add an additional requirement: true flexibility also implies the ability to use best-in-class hardware, software and services, where “best in class” means the ideal fit for your use cases and needs.
Put another way: If you can spin up hardware and software with ease but are unable to choose the solution that is best for your needs, that’s just flexibility with compromise. Why should you be forced to compromise? You should be able to pick and use the best RDBMS, the best big-data solution, the best operational data store, the best data visualization software, the best machine learning infrastructure, the best AI services….and so on and so forth.
Selecting a single cloud infrastructure provider and using the data services that provider offers, may deliver some level of convenience, but it also means a customer’s data is inextricably locked into that service. What if the provider isn’t meeting SLAs? What if there’s an extended service outage? What about a breach? If you’re locked in, your options for recourse are severely limited.
And how can you be sure you’re getting the best-in-class services I mentioned earlier? Going all in on a single provider is akin to splitting IaaS (Infrastructure-as-a-Service) and PasS (Platform-as-a-Service) options into 4 or 5 different groups, picking a single group and limiting yourself to the services in that group. Over the coming years public and private cloud providers, as well as independent ISVs, will compete fiercely to out innovate each other, and each will provide different best-in-class solutions. As a customer, you should be able to pick and choose the solutions that best fit your needs.
To really benefit from the flexibility of the cloud, you need to avoid lock-in and have the portability to move your applications across clouds or your data-center, with minimal friction and re-engineering.
Let’s take a look at a more pragmatic concern: cost. The public cloud allows users to “lease” hardware when needed. This is definitely cost effective for intermittent transient workloads which run for a few hours each day or week.
However, most mid-sized or large enterprises have mix of transient and sustained compute needs. While these demands may fluctuate, there is a certain amount of capacity that is required for the better part of the day. For such organisations, leasing compute from a public cloud provider at typical on-demand prices becomes cost prohibitive, particularly so for companies inclined to keep their datacenter hardware more than a couple years. For organizations that have sustained compute needs (a certain amount of compute capacity which is required for a significant part of the day), it may be more cost effective to:
- Run dedicated hardware in your data center
- Purchase reserved instances from your public cloud provider.
Reserved instances are pre-purchased from the public cloud provider for 1 to 3 year terms. These instances are available to the user 24/7, and the user agrees to pay for all of the hours of those reserved instances for the 1 to 3 year term. In return, the public cloud provider gives a discount on the pricing of those instances.
Reserved instances are a key component of lowering overall TCO in the public cloud. It is essential that organisations can purchase the cheapest reserved instances, wherever they are available. This will often entail negotiating with the cloud provider for discounts. Not being locked into that cloud provider’s infrastructure is a key benefit when you’re deep in price negotiations.
Obviously you lose the ability for pricing arbitrage if your data is tied up in proprietary applications. This is why application portability is so important. For most mid-sized and large enterprises, being locked into a single cloud provider, or to dedicated data centers, is unlikely to yield the lowest possible TCO.
In conclusion, for most mid-sized and large organizations, building applications that enable cloud portability is essential to leverage the flexibility and cost benefits of the cloud.
Cloudera provides best in class analytic database, operational database and data engineering software that can be run in your data center or on popular public cloud infrastructure providers, and can be ported with minimal effort. In future versions of our product, we will continue to improve our hybrid capabilities and make cloud portability as seamless as possible.