Migrating apps and other core resources to the cloud stands at the heart of many transformation strategies, but if the big three service providers pull the plug on access, where does that leave enterprises?
Parler’s recent banishment from AWS and re-emergence on Epik, a Russian domain registrar, is a very specific case but it does raise issues other organizations might need to consider -- especially in a market dominated by three hyperscale cloud providers. If one of those cloud providers permanently terminates services and other major providers in the United States refuse to take a customer on, what can companies do? Experts from CloudCheckr and Aiven offer some perspective on possible steps organizations might take under such circumstances.
Jeff Valentine, CTO of CloudCheckr, says while it is unlikely for most companies to face the exact situation as Parler, there are broader implications that organizations might want to consider. “There’s probably a really small percentage of people that will ever go through this,” he says. “But it should be thought about way before this happens.”
There are other reasons, such as sudden outages or the shutdown of a cloud provider, for organizations to create plans to salvage their code and get back online quickly, Valentine says.
Heikki Nousiainen, CTO at Aiven, also says the threat of getting cut off by all three major cloud providers is very low for most other businesses -- yet companies may want to maintain the ability to move code around for disaster recovery needs. “They are rare, but we sometimes see these big outages touch Google, AWS, or Azure in one or more regions,” he says. Companies with very time-sensitive online business needs, for example, may want to maintain the ability to roll over to a backup elsehwere, Nousiainen says.
He recommends exploring true multi-cloud options where companies can select providers freely without being locked in, and also going with open source technology because that lets the same set of services run in different clouds. Some of these options can come at a bit of premium, though Nousiainen says the overall benefits may be worth it. “There are costs associated but typically when that investment goes into preparing infrastructure as a code it also helps for many other problems such as disaster recovery.”
The design of an organization’s applications, Valentine says, can play a significant part in how companies proceed if cloud services are abruptly lost. “Most folks have applications that started on premise and then moved to the cloud,” he says. “That lift and shift strategy . . . has some drawbacks.”
It could be difficult to move the application again, Valentine says, because accommodations were already made to migrate to the cloud in the first place. If companies take the time to rearchitect applications, perhaps with a container model with Kubernetes, it can be more portable, he says.
There can be other pitfalls moving to the cloud, Valentine says, where organizations get locked in with a vendor because of the technology choices they made. To avoid such lock-in worries, organizations might choose technology such as open source that can work with many different providers to create a vendor-neutral platform, he says. “There’s a cost to that. Everything takes twice as long to build so it costs twice as much, so is it worth it?”
If a company were to face the imminent loss of its cloud services, Valentine says the only option is to take all of the code and implement any overdue changes in the moment. “There’s no way around it,” he says.
Despite the possiblity of losing access to cloud resources due to calamity or the cessation of a service contract, Valentine doubts organizations would suddenly revert to their old on-prem ways as a long-term alternative to operating in the cloud. “There’s no case for moving cloud to premise permanently,” he says. “I can’t imagine companies reversing course. Careers have been built on this digital transformation.”
Even so, Valentine says 20% of all apps are expected to always reside on premise though digital transformation is still in its early days as an industry trend. “We’re probably 10% of the way through this journey,” he says.
Potential changes that may make the cloud landscape more fluid could come from Google’s Anthos hybrid cloud platform, he says. AWS is also talking up run-anywhere technologies in its cloud, on the edge, or another provider’s cloud. “You end up having cloud vendors themselves acknowledging that cloud apps need to be more portable than they have been in the past,” Valentine says.
Other providers outside of AWS, the Google Cloud Platform, and Microsoft Azure are also coming into their own, he says, at layers higher than infrastructure. “Snowflake is the best example,” Valentine says, referring to the cloud-based data platform. “Instead of coding your own lake on infrastructure as a service using AWS, Azure, or GCP, you can buy the platform from Snowflake.”
If a company does find itself in a situation where it knows cloud services will be lost imminently, Valentine says they should consider these steps:
- Start a database backup of the relational database systems. “Those are likely to take time and you need a recent backup to restore somewhere else,” he says.
- Download object files to a local system. “You gotta get it somewhere, whether it’s temporarily into Dropbox, you just gotta get it somewhere else,” Valentine says.
- Assess the application code. If the organization has a continuous integration, continuous deployment pipeline of source code it owns but is in the cloud, this must be captured to be redeployed elsewhere, he says.
- Look at ways to change DNS settings. “It’s going to go into a blackhole when they shut you off,” Valentine says. If the organization can transfer to another registrar, he says, at least the company can point its users to an alternate landing page for the interim. “You can eventually redirect to somewhere else but you have to own your domain to do that, so you have to change registrars."
For more related content, follow up with these stories: