While it might be Australia’s largest telco, Telstra alone cannot compete on cloud computing prices and features when up against the scale of Amazon, Google or Microsoft.
On that basis, Telstra’s new deal with Cisco Systems looks to be prudent.
Under the deal, Cisco Systems will build and operate a cloud computing service inside Telstra’s Australian data centres that is based on Cisco hardware (servers, network) and Red Hat’s supported version of the OpenStack cloud computing framework (albeit with a few minor alterations by Cisco).
These racks will form a node in a global cloud Cisco operates with other telco partners across the globe – what the Californian company brands ‘The Intercloud’.
Just about everything in the stack is built and managed by Cisco – all features, functions, patches – despite it residing in Telstra’s data centre. Telstra concerns itself with network routes in and out of the data centre, customer support and billing.
It has big implications for Telstra’s IT staff, many of whom were laid off in a recent outsourcing drive, and even bigger implications for existing customers of Telstra’s Cloud Services.
Economies of scale
Erez Yarkoni [pictured], former T-Mobile CIO and now Telstra’s director of cloud services, spent time with me late last week to explain his strategy.
Yarkoni arrived at Telstra in November 2013 – his immediate concern being how the telco could achieve scale from its cloud computing investments.
Telstra’s initial foray into cloud computing – initially dubbed ‘Network Computing Services’ and later ‘Telstra Cloud Services’, was powered by the telco’s investment in hardware from EMC and Cisco, software licenses from VMware and BMC and an application migration and integration partnership with Accenture.
Several other Australian service providers – notably Alphawest, Bulletproof and Melbourne IT – had also built Infrastructure-as-a-service plays on the back of VMware’s hypervisor.
But at least two out of the three have since abandoned these projects as they struggled to compete with the global scale of Amazon Web Services, Google, Microsoft and others. The price war between AWS, Azure and Google in particular has rendered the most aggressive cost models from domestic suppliers uncompetitive to the point where there is more money to be made reselling and managing services hosted on some of these services than attempting to compete head on.
Yarkoni told iTnews he landed on the Cisco deal after a global technology tour that encompassed visits with most of the major players including HP, Microsoft and VMware. His discussions encompassed technology roadmaps but also “cloud economics”.
“Infrastructure is a global scale game,” he told iTnews. “And what defines the infrastructure is the software that brings it all together. I could try and compete by hiring 2000 developers in Seattle as Amazon and Microsoft has, or I could embrace open source and do it myself. But to do it yourself – to build your own set of software development resources – you are quickly going to find yourself unable to compete with global infrastructure.”
Yarkoni’s theory is that customers are likely to want to operate in a hybrid state.
“I’m of the same opinion as many Telstra customers – they will want to have more than one cloud infrastructure in play. The world will be a hybrid world. We will concentrate on building the service-managed infrastructure – such that we could, if required, attach Amazon Web Services or any other API-compliant cloud underneath.
“It’s important to understand what cloud economics mean,” Yarkoni said. “IaaS is a lean margin, global service. We need to graduate to building value on top of that infrastructure.”
Cisco’s Intercloud is built on technology the vendor already uses to manage its own WebEx Squared workloads, among others, allowing for a fairly aggressive roadmap.
Telstra already has racks being installed running the Intercloud software – specifically Red Hat’s version of OpenStack with Cisco’s management layer running over the top.
The next job is for Cisco and Telstra to build the necessary layers for the service to be commercially multi-tenant. By late May and early June, Yarkoni hopes to offer a platform in Alpha state, with customers being asked to “run some workloads and hammer the alpha environment” for stability while Telstra builds the billing interfaces to sit atop it.
In the latter part of 2014 Telstra will offer a beta service to customers for dev and test use, with the billing service now live, with full production ready before the close of the year.
Yarkoni agrees that its an aggressive timeline, but counts on Cisco and Red Hat to bring the necessary skills to the table to get the job done in a timely fashion.
I asked him if there were enough skills around OpenStack in Australia to do the job.
“That’s a great question,” he said. “Our aim is to build this with Australian hands, provide Australian support but use global R&D. I believe we will find the skills, if not [our partners] will import them.”
What does it mean for existing customers?
While the final deal is yet to be inked, its expected Telstra will have Cisco’s Intercloud exclusively in Australia for somewhere between one to two years.
Initially, Telstra intends to continue to offer its existing VMware-based ‘Cloud Services’ stack to customers as Intercloud builds up momentum. Yarkoni wasn’t keen to commit to its future development.
“We will respect all existing contracts, and support the features and functions of the last version [CSX],” he said. “And we’ll continue to actively sell it to customers that require dedicated hosting but want to run it on utility infrastructure.
“Over time, what I expect to be a journey of a couple of years, customers will either choose to – or we will help them – migrate to OpenStack with KVM or even with VMware if you have to. We’ll have all sorts of other flavours of cloud underneath.”
Yarkoni agrees that supporting multiple hypervisors or cloud architectures comes at a cost to Telstra. The beauty of OpenStack is that Telstra – and in turn its customers – will be able to choose whether or not to pay VMware’s software licensing fees or use a freely available option. Yarkoni said that if Telstra gets the service management layer right, what sits beneath will be less of an issue.
“If you walk into an Amazon data centre, you will find different versions of what Amazon considered their standard compute over time. But it all still works. Similarly, at some point, Apple decided to put Intel chips in laptops. It didn’t especially matter that you used a MacBook or a MacBook Pro, or that your computer came with one chip or another. Apple decided the economies of scale using Intel chips was better than designing their own.”
For that reason, cloud services built on Cisco’s Intercloud are likely to be branded ‘Telstra Cloud Services’ just as existing services are today.
A new service, and a new customer proposition
Yarkoni hopes that the Intercloud deal will bring about one key architectural difference between existing Telstra cloud services and those that will be on offer by the end of the year.
“Today, the elasticity we call ‘cloud’ exists only up to the edge of the data centre,” he said. “Where it needs to work is all the way to the edgepoint device where a customer consumes a service.
“What we have had to do in the past is to build a service layer that allowed us to address firewalling, routing, VPN, private networks, service instantiation, and all the move adds, and changes associated with those network issues. What we intend to build now is a service that automates this all the way to the customer.”
Telstra is also looking to incorporate network APIs (application programming interfaces) that allow a customer’s application hosted on the Telstra cloud to adjust the size and quality of the network pipe required.
“It would be ideal to say to customers: for certain apps, you no longer have to plan to a peak network capacity. You can have the app ask for capacity on the go. Or you can schedule a batch job to only run during off-peak rates.”
The Intercloud deal will – if customers subscribe at scale – elevate Telstra as a service management and IT operations channel.
“If we have the tools, and the people with a DevOps mindset, we can drive better economies of scale and operational excellence,” Yarkoni said. “We can bring the promise and elasticity of cloud without Telstra or the customer needing to over-invest in capital or planning for peaks.
“Then we can graduate conversations with our customers – how do you deliver that to my applications, the IT that supports my business? Sometimes that is a SaaS conversation, other times it is about bringing elasticity to the application – we pre-configure the infrastructure such that it becomes the click of a button in a catalogue.”