Over the last few years, we have seen Oracle go from being a cloud denier to proclaiming itself the largest cloud company (by some measure or other). I had not been exposed to the Oracle cloud message until very recently, so I was cynical about what I heard. One challenge is that Oracle is known for lawyers, aggressive licensing, and marketing to the CxO level. None of these are things that resonate with me, so I knew very little about Oracle’s products and particularly its cloud strategy. In May, I spent some time being briefed about Oracle’s new cloud and came away impressed with its future. The event was a blogger briefing by Ravello Systems, which was acquired by Oracle in February 2016. I was one of thirty bloggers invited to spend a day at the Oracle offices in San Francisco. Nearly half of the presentation time was spent on the Oracle cloud. I will write about the half that was spent on Ravello in another article.
Disclosure: Oracle/Ravello paid for my travel to San Francisco, hotel accommodation, and some nice food and drink. Oracle/Ravello did not commission this article or have any oversight before it was published. I have done paid work for Ravello in the past and hope to do more work with it soon.
Oracle Cloud History
As it is an application software company, it is no surprise that Oracle built its cloud for Software as a Service (SaaS) first: essentially, the ability to have Oracle databases in the cloud with some supporting software, also as services. The reality of how databases are consumed is that there needs to be code execution close to the database. Oracle added Platform as a Service (PaaS) functions. It discovered that PaaS isn’t the best solution to every coding problem, so it needed to add Infrastructure as a Service (IaaS) VMs to the product set. It seems that this evolution did not lead to an optimal cloud platform, and Oracle made a hard decision. It chose to build a new cloud platform that will be good for IaaS and on which it can build PaaS and SaaS platforms.
A New Story
It does appear that Oracle is building its second-generation cloud platform to be an enterprise-class cloud platform. What I mean is that it will allow customers to bring their workloads in without having to completely rearchitect around the limitation of the cloud platform. This is in stark contrast to AWS, where the infrastructure is best suited to twelve-factor applications. This Oracle Cloud lets you have bare-metal servers (and even install your own hypervisor), VMs, and PaaS or SaaS, all delivered in one location. To provide the bare metal, Oracle has the network virtualization in the network, rather than in the hypervisor like AWS. Customers can even run their own PXE services for operating system deployment. Oracle also offers high-performance storage without needing vast numbers of disk devices attached, up to 200 K TPS for an NVMe-based server. Everything seems to be focused on delivering service levels and capabilities that are in line with on-premises enterprise data centers.
The enterprise features include the ability to have administrative subdomains within an organization. Rather than using a credit card, a development team might have control of its own cloud infrastructure, with a budget limit, while corporate pays the bill. Speaking of billing, it sounds quite unlike my expectation of Oracle. The pricing is supposed to be simple, and discounts are given for total spend, rather than being tied to a specific product as AWS does with reserved instances. There is more that seems new for Oracle; it is committed to using HashiCorp Terraform as its Infrastructure as Code platform, rather than building its own tool set. One great customer benefit of Terraform is that it can be used to orchestrate provisioning and configuration across multiple clouds: no lock-in. Oracle seems to be happy to minimize lock-in; its network transit and egress costs are low. There are some normal parts to its new cloud architecture: availability domains that correspond to one or more data centers, and regions made up of interconnected availability domains in the same geographic area. However, this is a differentiated cloud, not a direct competitor for AWS, as it has different objectives. Oracle’ second-generation cloud platform has characteristics that make it friendly for existing enterprise workloads.
This new cloud is still in its very early days. Right now, there are only two active regions, both in North America. There are a couple more regions in the near-future roadmap. Hopefully, some Asian regions will be added as well. It is great to see Oracle building a public cloud that fits with its customer base. It does remain to be seen whether Oracle will make a success of a public cloud platform. HPE, Cisco, and VMware have all withdrawn from providing their own public clouds in the last couple of years. It is a hard problem to solve, and a place to spend an awful lot of money. Time will tell whether Oracle has made the right choices and has the resolve to build success in the long term.