CIO

Blueprint for the next generation data centre

Can another layer of software-based complexity really help win the decades-long war on infrastructure management costs?

Given the speed with which Australian organisations have virtualized their x86 server architecture, it’s not surprising many are investigating which other parts of the data centre are ripe for abstraction.

Infrastructure virtualization has moved well beyond the humble server to now take in storage and networking infrastructure, and thanks to the emergence of copy data tools such as Actifio, can even extend to the virtualization of an organisation’s data.

The concept of virtualizing everything is the foundation of the new hardware category of converged infrastructure, as embodied in VCE’s Vblock Systems, which bring together skills in virtualization, networking and storage from VMware, Cisco and EMC respectively into a single box.

The end state for virtualization is the software-defined data centre (SDDC), where all functions are managed at a software layer abstracted above the hardware, enabling complex functions to be deployed on commodity hardware.

While that vision is still some way from realisation, many organisations are working towards its partial fulfilment through the creation of hybrid cloud environments. These are intended to deliver the management and resource utilisation benefits of the cloud within an organisation’s own data centre.

But can another layer of software-based complexity really help win the decades-long war on infrastructure management costs? And are Australian organisations ready for software-defined everything?

Supporting agility

One organisation that is willing to find out is Woolworths. As one of the 15 largest retailers in the world, Woolworths has massive revenues but operates on razor-thin margins. That presents a challenge for head of infrastructure, Matt Chamley, whose role requires balancing that restriction against other business needs.

“One of the things I’m challenged to do as the head of infrastructure is to support the business to have agility, and by that I mean flexibility and responsiveness to changing dynamics in the market,” he says.

“So how do I get infrastructure off the critical path? That’s the problem statement I’ve been working with.”

Hence his interest in virtualisation. Woolworths is progressively moving workloads off existing reference architectures and on to two Vblock boxes, which now run services such as the retailer’s Citrix desktop deployment.

Chamley says the benefit of Vblock is that he and his team can be very ‘hands off’ in terms of its management, as all changes and upgrades are managed by VCE.

“With engineered platforms and true converged infrastructure, such as the Vblock, it allows us to abstract our thinking away from the componentry,” Chamley says. “We know there is a capability and a pool of resource that we’ve bought, based on certain performance metrics, and that it is designed to work. So we can now start looking at how that service is being consumed.”

If Woolworths follows through on that vision, the creation of a single pool of resources could be far-reaching for Chamley’s team, as emphasis moves from managing the old model of ‘data centre, core, edge’ in favour of managing the services delivered.

“If I could manage workloads based on characteristics and automate the way those applications are defined and deployed, then all of a sudden I don’t need a team of infrastructure engineers anymore,” Chamley says.

“I need a team of infrastructure developers or architects or business consultants.

“Ultimately the service we provide the store is its ability to trade and its ability to recover from an outage of trade. And we are starting to define our services based on those business outcomes.”

That is the long-term vision, but Chamley is also getting short-term benefits in the form of reduced management overheads, particularly for systems updates.

This has been demonstrated during two major updates to the Vblocks installed under the Vblock Release Certification Matrix program.

“Twelve months ago, the service owners in that space wouldn’t sleep for days, people would be working weekends, and it would probably take four to six weeks to do after the testing and validation had been done in the lab,” Chamley says. “Today the service owners just go to sleep and the next morning it’s done.”

He is also achieving his goal of greater responsiveness to business needs, and can now create new platforms in days, rather than months.

Fully virtualised

That notion of responsiveness is also a driver for the ongoing virtualisation program at Westpac New Zealand.

Principal enterprise architect, Anton Aalders, says the bank will move into a new data centre in 2015 which will be almost entirely virtualised in terms of both computing and storage. Westpac NZ has also virtualised its data using Actifio, enabling faster staging of testing services. The next area of investigation is software-defined networking using Cisco Application Centric Infrastructure.

Aalders says the march to virtualisation was driven initially by the desire to move from legacy data centres without dragging across aging infrastructure. The new data centre will operate as a private cloud, with pattern-based deployment and automation.

“The business case was really ‘let’s not buy 50 boxes, lets buy three or four’,” he says. “And we wanted to get some agility, we wanted to have the capability to ratchet up, ratchet down, and resize.

“More and more people are starting to understand a cloud isn’t a managed service, it actually is something that is pretty agile and pretty flexible, and they are starting to want some of that. And the legacy IT guys are finding that if they can’t deliver it and don’t do it for their organisation, the organisation is going to do it around them.”

However, Aalders says the true benefit of virtualisation in terms of business agility is not always easy to measure.

“Sometimes it is hard to quantify the agility, especially when the technology largely can do it, but people can’t adopt the process to get the agility,” he comments. “We are driving that into Digital [the bank’s digital division], but ultimately we want that on bank side, so everyone becomes agile in their methodology.”

Next up: Services centricity

Page Break

Services centricity

Like Chamley, the push to virtualisation is changing the way Aalders’ team works, as they shift beyond managing infrastructure to managing services. “They no longer just do an application or a database or a technology, but they have become service-centric,” he says.

For Melbourne-based RMIT University, the next step in the virtualisation journey involves the ongoing virtualisation of its access network infrastructure using OpenFlow-based technology from HP.

RMIT’s deputy director for infrastructure delivery within its Information Technology Services team, Ben O’Neil, says the university has witnessed a 700 per cent increase in BYO devices coming onto the network in the past two years.

“That exponential growth that we are seeing in the field has put a lot of stresses on systems that typically are very expensive to scale,” he says. “What we have started to do is look to SDN to programmatically apply those services at the access network level.

“A lot of RMIT’s IT transformation over the past few years has been user-centric. So when we looked at SDN we looked at how it could solve and improve the end user experience – their performance, their access, their security.”

The move to SDN is allowing RMIT to better manage its network traffic, as traditionally many of the packets passing through the core received no added value, but generated a high cost overhead through the resources used to move them around.

“What we are really wanting to avoid is needing to do large-scale repeated upgrades of our core infrastructure,” O’Neill says. “That equipment already equates to 16 per cent of the IT infrastructure costs at RMIT, so if we need to expand that 700 per cent you are talking a lot of money, and that percentage would grow substantially.”

O’Neill says managing network traffic in this way should also speed delivery of services, as packets will traverse less infrastructure, and allow for the reduction in size and capacity of some firewalls. The results could be significant.

“The expectation I’ve set with the IT steering committee has been that we think we can pull out a million dollars a year in running costs,” O’Neill says. “I would expect it’s the exact same business case in many places.

“If you can’t explain the benefits to a CIO in eight minutes you are doing something wrong – it is that straight forward.”

RMIT began deploying HP OpenFlow switches in April 2012, with 550 seats now deployed and another 2000 expected to be added by March next year. The goal for 2015 is to have 10,000 of the university’s 20,000 swapped over by the year’s end.

The university is also writing its own SDN applications in partnership with its School of Electrical and Computer Engineering, and is encouraging students to develop for the network.

“They were super keen to take that on and the lecturers were very supportive,” O’Neill says. “This year, they are very keen to start taking this into the curriculum.”

O’Neill says the process of rolling over to OpenFlow has been relatively simple. OpenFlow-enabled switches and routers are simply configured in a traditional fashion at implementation and then switched over to the SDN controller one port at a time.

“You can leave the legacy configuration in place and tell that port to be driven by the OpenFlow controller,” O’Neill explains. “If the controller fails for any reason or of the flows aren’t working, or something is happening that you can’t quite get to the bottom of, it’s quite easy to turn the controller off and let the legacy traditional network configuration take over in an instant.”

But will redefining the data centre in software ultimately lead to reduced costs, and finally allow IT to get on with assisting the business, rather than maintaining itself?

For Woolworth’s Chamley, the ongoing virtualisation of his infrastructure holds the promise of making this possible.

“For me it is about getting off all hardware, getting off all tin,” Chamley says. “I don’t want to be constrained by a data centre. I need portability of my workloads. And three or four years ago that may have been from one data centre to another, but now it includes public cloud providers. And that’s where the flexibility comes in.”

“As we head into what I’ve been calling infrastructure 3.0, it’s all about abstraction. It’s about the orchestration tool sets which are no longer technology-specific, but enable you to manage the entire infrastructure substrate and give you automation.

“And ultimately the net result of infrastructure 3.0 is getting infrastructure off the critical path.”

Virtual data?

What do you do when you run out of hardware to virtualise? You virtualise the data itself.

At the vanguard of this emerging capability is the US-based developer, Actifio, whose technology virtualises the so-called copy data (multiple copies of the same data in an organisation) to create a single maser copy that can then be used for activities such as disaster recovery and test and development.

This has the benefit of both significantly accelerating the creation of new environments while reducing data storage requirements and creating a more reliable backup regime.

Sydney-based law firm, Henry David York (HDY), is using Actifio to virtualise its data environment, having already virtualised almost all of its server and storage infrastructure.

Its manager of business technology solutions, Alex Clonaris, says the use of a copy data tool has significantly improved the reliability of HDY’s data recovery strategy, and enabled faster staging of test & development environments.

“Staging and testing take up production time,” Clonaris says. “By having multiple environments available to be spun up dynamically, you speed up the time it takes to do testing and move on to the next iteration.

“What that means is we’re able to speed up the delivery of applications to the business.” That aligns well with Clonaris’ goal to reduce the time HDY invests in managing infrastructure and focus more on using IT to deliver value for the business.

“What we’re investing a lot in now is the concept of the business technology solutions piece,” Clonaris says. “It is a consolidating approach, working with each group and trying to refine their processes and what we can do to help them. That is where we see the biggest value moving forward.”