IaaS in the Optus “virtual private data centre”

Optus has unveiled its cloud computing service, dubbing it a "virtual private data centre."

It’s not about the public cloud: with its launch yesterday of a new cloud computing service to go live on October 1, Optus is explicitly targeting the world of the private cloud.

Foreshadowed in the first quarter when Optus announced Curtin University as a pre-launch user, the service is based on a simple infrastructure-as-a-service (IaaS) model: customers of Optus’ MPLS-based Evolve IP Network will be given access over their private networks to infrastructure in the Optus data centre.

Acting managing director of Optus Business Rob Parcell said the IaaS offering is designed to target customers’ wish to access the cost effectiveness and flexibility of cloud computing solutions.

Citing IDC research, Parcell said the top-rated cloud computing services are those that provide access to servers and storage, with e-mail, data backup and storage the workloads moving into the cloud soonest.

In creating this private cloud IaaS offering, Parcell said Optus is “taking advantage of IT infrastructure that’s close to the network”. Delivering the service over the private MPLS network makes it easier for Optus Cloud Computing to address the security and performance concerns that are holding back cloud implementations among enterprise customers.

Instead of a public cloud, the Optus Cloud Computing IaaS service is a “virtual private data centre”, Parcell said: “Optus can provide an enterprise cloud that runs on the same network as the customer’s system.”

The service is designed to be both flexible, within the simple set of features offered at launch. Customers purchase IaaS access as discrete units of processing power and memory: each 0.5 GHz worth of processing with 1 GB of memory will cost $125 per month, with two performance increments of storage at either 30c or 60c per GB per month.

Once signed to the service, customers will be able to raise (or delete) processor instances themselves on an on-demand basis.

Optus will offer pre-packed Window-based server install, with customers able to self-install other operating systems. The environment supports any operating system that can run under a VMWare virtualisation environment.

While the feature set looks limited, as Andrew Vranjes (data centre practice manager at Alphawest) points out, it “empowers users to create their own virtual infrastructure for their own use-case.” Instead of constraining customers to selecting those features enabled by Optus, customers can build their own systems on the Optus infrastructure.

It’s about the SLA

They key pitch of the Optus IaaS service is its SLA support. A public cloud can only offer SLAs over the infrastructure it controls – access to the system is subject to Internet connection outages over which the cloud provider has no control.

The private network delivery of Optus Cloud Solutions means customer access can be managed all the way from the managed router at the customer site through to the data centre. The network is covered by the Optus Evolve IP SLAs (covering network availability, latency, jitter, packet loss, restoration requirements and so on), while the IaaS carries a four-nines availability SLA. Customers will be rebated against the IaaS SLA, as they would be if the network failed to meet its SLAs.

The exception to the private network access model is where customer organizations need to provide Internet-based access to the applications they have deployed to the service (for example, in the case of teleworkers). However, while end users will be able to interact with their company’s applications from the public Internet, they won’t have access to the administrative capabilities.

Vranjes said a use-case such as this illustrates the dividing line between making the “cloud” available over the private network, and on the other hand making the services available over the Internet.

The browser-based management interface to the service is divided into three constructs, covering resources (individual virtual server instances), content and applications (deployed to the server instances), and users / groups.

Vranjes said one of the key reasons customers are looking to move their virtual machines (VMs) into the cloud is that many large customers are experiencing “virtual machine sprawl” in which users or administrators create scores of more-or-less ad-hoc virtual machines which persist (and consume resources) even when they’re no longer in use.

The management system allows administrators to set an expiry for a new VM instance, so a machine can be created for a temporary purpose such as application testing, without users having to remember to destroy the VM when it’s no longer needed.

Read more on Cloud computing services