leungchopan - Fotolia

How Oracle differentiates its IaaS against AWS

Cloud computing is changing both IT infrastructure and software development. We take a look at Oracle’s plans as it starts to compete with Amazon Web Services

Oracle is now ramping up its public infrastructure as a service (IaaS), the Oracle Bare Metal cloud.

The company recently held a cloud developer day at CodeNode in London, where senior executives and its cloud technologists discussed the merits of its IaaS platform over Amazon Web Services (AWS).

Oracle’s second-quarter 2017 results show that its cloud revenue grew by 13% to $1.2bn. Commenting on the results, CEO Larry Ellison said: “Our second-generation IaaS business is both faster and lower-cost than AWS. And now our biggest customers can run their largest and most demanding Oracle database workloads in the Oracle Cloud – something that is absolutely impossible to do in the Amazon Cloud.”

The company appears to have a two-pronged approach. At the top end, Oracle Fusion and its software-as-a-service (SaaS) products are the paths Oracle wants senior IT leaders to consider as they upgrade their on-premise Oracle systems. Costs can quickly add up to multimillion-pound annual subscriptions.

At the time of writing, its financial ledger SaaS costs £473 per user per month, but given that it is licensed for a minimum of 10 users, the annual cost is more than £500,000 for 10 users.

The second prong of attack is targeted at building a groundswell among the developer community. Here Oracle claims its IaaS platform is 70% cheaper than AWS.

Deepak Patil, vice-president of product development at Oracle, believes that, over time, the cloud will be the primary infrastructure on which most IT workloads will run.

Logically, this means applications developed natively for the public cloud will eventually surpass on-premise applications.

“There are 60 million x86 servers running currently,” said Patil. “Out of these, 7% are shipped to the public cloud, which essentially means 7% of workloads are running on the public cloud. Within five years, 40% of servers will ship to public cloud providers.”

Oracle is setting itself up as one the major public cloud providers, alongside AWS, Microsoft, Google and IBM.

It has more than 20,000 people working on its cloud, and has reportedly hired many from AWS. Patil was one of the founding members of Microsoft Azure and is now responsible for helping Oracle build its next-generation cloud platform.

Oracle’s vision is Cloud 2.0, to tackle some of the problems it sees inherent in the architectures adopted by AWS and Azure. In a recent London presentation for developers, Patil highlighted networking as the key differentiator for Oracle.

Network differentiation

He said Oracle is building a flat network with capacity. It is providing low-latency, high-bandwidth interconnectivity between availability domains, encrypted over fibre. Its cloud comprises regions, each of which has three fault-independent availability domains connected by different electrical grids.

According to Patil, in the Oracle cloud there is one millisecond latency across multiple datacentres within a region. “Network ports was the limitation in my previous life,” he said.

Speaking about the core engineering decisions Oracle has taken, Patil said the Oracle cloud has predictable low latency, with less than 100 microseconds latency across nodes within a region.

Another of Oracle’s engineering decisions relates to virtualisation of the network. Rather than run networking on top of the hypervisor (as in a software defined network), which is how public clouds usually operate, “everything runs in a private network layer”, he said.

Oracle heats up competition

Oracle has cloud regions on the East and West coast of the US, along with separate regions for the Department of Defense, the US government and US healthcare. In the UK, there are currently North, South, and a UK government region. Oracle also has an EU region and a separate region for Germany. Asia-Pacific, Australia, New Zealand, Singapore and the Middle East are also covered.

It has built a partner ecosystem with more than 700 open source providers to support its cloud. Canonical, which distributes the Ubuntu Linux operating system, is among these. Its CEO, Mark Shuttleworth, was invited on stage by Patil at the recent London developer event, when he said: “Business today is software-defined. If you got a big cheque, you would end up with the Oracle cloud. It represents proven state of the art.”

Owen Rogers, research director at 451 Group, said: “Oracle is really going for this. It is miles behind AWS, Azure and even Google, but it has the money to make R&D investment and can take risks.”

Oracle recently changed its pricing in what appears to be a bid to make it less attractive for its customers to run the Oracle database on AWS. In a briefing note about this change, Rogers wrote: “When running on AWS, Oracle stated that each AWS virtual CPU (vCPU) should be licensed as half a core. It recently changed its agreement to licensing each vCPU as a whole core, doubling the price of running select Oracle software on AWS. The Standard edition database is unaffected, but the Enterprise edition’s price will increase.”

Look at other options

The 451 Group expects some users could move their databases to the Oracle cloud as a result, but others may look at other options. “AWS has a range of cloud databases, and although migration won’t be cheap or easy for enterprises, a doubling of price might make it worthwhile,” Rogers said in the briefing note.

One organisation that has moved from AWS to Oracle is AnzenData, which specialises in data jurisdiction. Its core technology separates data that needs to be kept in a specific geography, such as anything that can personally identify someone and collective data (where people cannot be personally identified) that can be stored in a public cloud. The company is working towards being able to move gigabits of data per second with a latency of less than 500 microseconds, in a bid to move and recluster data reactively.

Gary Mawdsley, CTO at AnzenData, said: “We have been running on AWS, but last autumn we came across Oracle Bare Metal cloud. If I launch a machine on Bare Metal, I don’t have to worry about the network. Under Oracle Bare Metal, the networking is seen the same way whether we run virtual machines or bare metal servers.”

With AWS, it is necessary to make configuration choices, said Mawdsley. “I have to fiddle around to find which machines I should be running and the AWS machines are specified in certain combinations, which implicate network capacity so you have to make a choice of which machine to use. On Bare Metal, you don’t.”

Bandwidth customisation

Looking at Oracle’s bandwidth differentiator, 451 Group’s Rogers said: “Google has custom sizes to enable users to specify exactly what they want. AWS has different virtual machines and stipulates high, medium or low network capability for any VM. They are constrained by how much you can bring in and out of the cloud, but you have to choose from fixed sizes.”

But, in general, virtual machine sizes grow linearly, he said. “If you have a small amount of compute and storage, then you will probably also have low bandwidth. And as you grow, your bandwidth grows.”

What this means on AWS is that a workload with high compute requirements is likely to require high bandwidth, but something running on AWS’s T2 instance, which is designed to provide moderate, baseline performance, would usually have a lot smaller bandwidth requirement.

Rogers believes that building applications in a cloud-native fashion may water down Oracle’s somewhat compelling differentiator. “If you are building a cloud-native application, you should use really small virtual machines and code in a way that adds more VMs as the workload grows.”

Even though a small VM would have a small amount of compute, storage and network bandwidth, each additional VM grows the pool of these resources, which enables the cloud-native application to grow in an elastic manner, said Rogers.

Read more on IT architecture