The OpenStack summit in Barcelona on 25-28 October is expected to be the foundation’s biggest-ever European event, and reflects the growing maturity of the open source cloud.
“Our 14th summit is the second one in Europe and we are expecting 6,000 delegates,” said Jonathan Bryce, executive director of the OpenStack Foundation.
Speaking to Computer Weekly during an OpenStack Day in Bristol this week, Bryce said: “We have had a lot of growth in the use of OpenStack.” For example, he said Sky UK had built an OpenStack cloud to support business applications. Santander is also a major user of the technology.
Bryce said Europe was experiencing huge growth in public clouds. There is an HMRC cloud in the UK and in Sweden, there is a city cloud supporting insurance companies and banks, he said. Both of these are built on OpenStack, providing country-specific public clouds.
OpenStack’s fastest growth area has been telco workloads using network function virtualisation (NFV). “Most of the big telcos have OpenStack in their stack,” Bryce added.
The next software update of OpenStack – Newton – is being released just before the October summit.
Bryce said bare metal provisioning was among the new features in Newton. “You can eke out a higher level of performance or use specific hardware, such as GPUs,” he added.
OpenStack was also introducing a new programme for professional certification and workload-specific architectures, said Bryce, which reflected a greater level of maturity and helped OpenStack to become more enterprise-ready. “We have reached first level of maturity in the last year or two,” he pointed out.
Read more about OpenStack
- People are no longer just dabbling in the OpenStack private cloud – the OpenStack Foundation has found many users are starting to undertake production work.
- This guide looks at how OpenStack Cinder, OpenStack Swift and OpenStack Manila handle storage, and weighs up how OpenStack compares with VMware.
Bryce accepts that the biggest challenges now facing OpenStack are cultural rather than technical. In many ways, the maturity of OpenStack to become more enterprise-ready had followed a similar evolutionary path to that of Linux, he said.
“In 1994/5, Linux was a very technical operating system. But in the late 1990s to 2001, we started to see commercial efforts to make it accessible and trusted by enterprises,” he added.
But in OpenStack, physical infrastructure was often replaced by software-defined IT infrastructure, he said. “What you do with Windows and Linux administrators is the bigger challenge for a lot of companies. Once you prove the technology, you need policies and training to push people.”
Bryce said one of OpenStack’s banking customers had established a corporate policy of all physical hardware needing board-level approval. This was backed up by training to show staff the benefits of self-provisioning on OpenStack.
Paddy Power Betfair, formed from the merger of bookmakers Paddy Power and Betfair in 2015, is using OpenStack to update its physical infrastructure in a way that can scale easily.
Two years ago, BetFair CTO Paul Cutter realised that the company’s infrastructure would need to be refreshed over the next two to three years. “We went out to the market to look at what could be our next-generation infrastructure, and this became the start of our OpenStack journey,” Richard Haigh, head of delivery enablement at Paddy Power, told Computer Weekly.
“We decided to de-risk our decision a little by using Red Hat’s distribution of Open Stack.”
When BetFair began asking for requests for proposals, it did not know exactly what it needed, said Haigh. “To be fair, we didn’t have all the answers or all of the questions. But neither did the suppliers. We knew we needed a partner who understood what we were doing.”
Among BetFair’s challenges was identifying best practices gleaned from existing customers.
Specifically, said Haigh, it was very hard to find reference customers using software-defined networking on OpenStack at the scale, automation and control that BetFair needed.
It wanted to operate OpenStack at 500-600 nodes per datacentre, but Haigh said: “We struggled to find customer references at the node size and transactional throughput we wanted at BetFair and we were worried that at above 200 nodes, Neutron [OpenStack’s NFV] would be less performant.”
The SDN would become a key aspect of the company’s ability to scale by routing traffic dynamically, rather than adding more physical application servers, he said.
While many telcos use OpenStack for NFV, Haigh said their approach was very different from what BetFair was looking for. “Their infrastructure tends to be quite static, while infrastructure is much more dynamic and things at BetFair move far more quickly than in a telco world,” he said.
Also, BetFair desperately wanted to avoid supplier lock-in, something telcos are probably happy to deal with, given the massive scale they generally need to run at, said Haigh. “We wanted to put everything through the OpenStack API (application programming interface) as this gives us the flexibility to change when we want.”
Looking at the enterprise challenges described by Bryce, especially the culture change, Haigh said that moving to OpenStack and NFV had not meant getting rid of the company’s networking team.
“They still have a job,” he said. “For me, the network is split into two halves – one is under software control and the other is the edge networks and the wide area network and the connectivity that is not under software control, which all need to be maintained. We still need guys who can build and understand these things.”
While many CIOs may feel their organisations have little in common with online bookies, Paddy Power Betfair’s experience illustrates the corporate challenges all organisations will face as they look to deploy OpenStack.
Clearly, choosing a commercial distribution, such as Red Hat, gives enterprise-level support and, as Haigh said, “de-risks” the project. Among the key take-aways is the fact that there is no such thing as an “OpenStack solution”.
The company’s implementation showed that it needed to work closely with Red Hat – and neither the customer nor the supplier truly understood what the final outcome needed to be.