Check out the on-demand sessions from the Low-Code/No-Code Summit to learn how to successfully innovate and achieve efficiency by upskilling and scaling citizen developers. Watch now.
When cloud computing first became widely available, the advantages were clear: ease of use, unlimited capacity, availability, and flexible pricing. Suddenly, enterprises didn’t have to worry about capacity planning or the cumbersome and expensive process of setting up data centers. Cloud vendors empowered organizations to focus on building their products and their core business instead of setting up and maintaining costly infrastructure, allowing them to scale at an unprecedented pace.
These advantages led to the rapid migration of many enterprises, startups, and new businesses, from on-premises to the cloud.
With time, however, the costs associated with relying totally on the cloud gave rise to new problems. Overprovisioning of cloud resources is one of them. Overprovisioning is a standard practice from the days of on-prem computing that has made the leap to the cloud, primarily as a result of companies employing “lift and shift” strategies for migrating, and it has sent cloud costs skyrocketing. As new companies got off the ground and looked to begin achieving profitability, budget-conscious executives naturally scrutinized the hefty sums devoted to cloud spending and began asking whether all that money was generating sufficient ROI.
Dropbox’s decision to abandon AWS and build its own network of data centers, which famously saved the company $75 million in just two years, continues to resonate. But not every company is Dropbox, and on-prem carries its own costs and complexities.
Intelligent Security Summit
Learn the critical role of AI & ML in cybersecurity and industry specific case studies on December 8. Register for your free pass today.
As companies globally began to prioritize reducing infrastructure costs, many began to contemplate the advantages of the hybrid model. This approach, which relies on on-prem infrastructure and uses the cloud to scale out for peak traffic, is poised to strike the perfect balance.
Two industry trends are making this possible: managed services offerings launched by cloud vendors such as AWS Outposts, Azure Arc, and the Google Cloud Services Platform, Anthos, which allows for dynamic auto-scaling when necessary.
While it would be tempting to view this model as the best of both worlds — the ability to both run off on-prem infrastructure and benefit from cloud resources — and an easy decision to make, hybrid deployment comes with its own unique considerations and challenges. Here are some key points to consider before taking on a hybrid approach for your company.
1. Determine strategy and priorities
Not every enterprise needs to bring its computing infrastructure in-house, but companies that have flagged cost reduction and margin growth as strategic priorities should explore doing so. Likewise, enterprises that are less concerned with their margins at the moment and aim to scale and rapidly increase their market share can comfortably stay fully in the cloud to maintain a greater degree of flexibility.
2. Prepare for the realities of on-prem maintenance
A hybrid strategy means a return to some of the complexities of on-prem infrastructure and management that enterprises left behind when they blasted off to the cloud. These challenges should by no means prevent a company from going hybrid, but they do require a carefully planned on-premises strategy. The capacity to take on that challenge may be influenced by your institutional memory for self-management. It is essential to have staff who know how to manage data centers, procure servers, and so on. Enterprises returning to on-prem after only a year or two away will be at a clear advantage — they may even still own their facility. Cloud-native organizations will be starting on that journey from scratch and would benefit from bringing in people familiar with self-managed infrastructure.
3. Know, and constantly re-evaluate, your triggers
A hybrid approach is all about planning for the usage threshold at which you scale your application out to the cloud. That necessitates effective capacity prediction. A general rule of thumb would be to plan your on-prem for average traffic, not peak, and scale-out to the cloud when experiencing peak traffic.
4. Get comfortable with fragmentation
Scaling out from your metal to the cloud and back again is no mean feat from an infrastructure perspective. But it can also stress and expose aspects of your application itself. If parts of your app are simultaneously running in multiple areas, you have to ensure that your data and code base are uniform across each. To put it plainly, think about the cloud as simply an additional data center — you’ll need to guarantee constant data updates to ensure consistency.
5. Consider the lock-in
It may seem counterintuitive, but using the cloud on top of data centers purely to service excess demand actually comes with its own kind of hook. Instead of having the option to mix and match data center and cloud vendors, the major cloud vendors’ managed services solutions for running in data centers scale-out exclusively to their own clouds. The vendor choices for hybrid are just the same as choosing a pure-cloud vendor. Consider whether you can be tied to a particular API and how large the ecosystem is. Identify the specific features that are priorities for your businesses, as some are only available from certain providers.
6. Let location guide you
Faster 5G speeds for consumers may reduce the need for enterprises to keep complex networks of regional services. But the value received from doing so depends on your user footprint, as well as 5G’s roll-out roadmap. In theory, if you serve customers only in the US, enhanced 5G speeds may allow you to jettison those “east” and “west” cloud regions,# and instead consolidate onto a single US data center. But a distributed customer base may still demand multi-regional power that the cloud is best positioned to provide, while 5G’s still unrealized rollout means uncertainty abounds – and will for several years to come.
Hybrid: The next frontier
Over the next five years, hybrid infrastructure deployments are likely to become increasingly commonplace as many businesses reach a point at which cost-saving becomes imperative and the barriers to entry continue to break down. As cloud vendors’ managed-services APIs becoming more user-friendly, hybrid will emerge as a dominant go-to solution.
Asaf Ezra is CEO and Co-Founder of Granulate.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.