The diversity in the open source software world can be a boon and a bane to wider adoption in the enterprise.
After all, without the right knowhow, it can be hard to figure out how they are going to work together on existing infrastructure – and if the chosen projects will eventually survive.
That’s where open source companies such as Suse step in. While smaller than US-based rival Red Hat, Suse has found its footing in identifying and supporting open source projects that help to run mission-critical enterprise workloads, improve developer productivity and solve business problems in industries such as retail.
In an exclusive interview with Computer Weekly, Thomas di Giacomo, Suse’s president of engineering for product and innovation, talks about the company’s efforts to make open source software more palatable to enterprises, the state of open source software adoption in Asia and his views on IBM’s recent acquisition of Red Hat.
How has the open source model evolved over the years? Would you say the diversity of open source software is also its weakness in the enterprise context, as many companies in Asia still find it hard to stitch together different open source projects to address a business problem?
Thomas di Giacomo: Open source today is a lot different than what it was 25 years ago. When Linux first started in 1991, it was about doing what non open-source software was doing it in a cheaper way. But now, all the innovation is in open source and the likes of Microsoft, Google and Facebook all innovate in open source.
The challenge for enterprises is that they have a business to run and avoid being disrupted. And open source can help them, but it’s very difficult for a company to decide which kind of open source projects to use for their needs.
Then, as you mentioned, even if you pick up one open source project here and one project there, nobody’s telling you how they are going to work together on your existing infrastructure, and if those projects will survive.
Today, there are thousands of open source projects. You’ve got so many foundations, such as the Linux Foundation, OpenStack Foundation, Apache Eclipse and the Cloud Native Computing Foundation where Kubernetes is hosted, for instance. There are also thousands of projects like TensorFlow that sit outside those foundations.
What a company like Suse is doing is to help enterprises such as banks, healthcare providers and retail companies match what they’re trying to do with what’s available in the open source world. We select the projects and make sure they can work together with enterprise IT infrastructure, and are stable, secure and supported over time. We’ve started doing that with Linux, OpenStack, Cloud Foundry and Kubernetes.
Now, you mentioned Asia. The challenges I mentioned are common to everybody, but what we see in Asia, like in Europe, is that Asia is not a single, homogeneous market. Different countries are in different stages of adopting open source. I spend quite a lot of time in Japan, China, Hong Kong, Singapore, all of which are very different markets.
Typically in Japan, enterprises are more conservative so we have a lot of customers like banks that are running Linux on mainframes. Singapore is more innovative, so we see OpenStack being used by the public sector and manufacturing companies.
In Taiwan, we have projects that run AI (artificial intelligence) applications on top of OpenStack as well. At the operating system level, Linux is dominating and now we see other open source projects picking up for private clouds and containers.
What do you think are the biggest barriers to greater adoption of open source software in the enterprise? Today, the lines of business have a seat at the table when it comes to making technology decisions. Is there a need for a change in the way open source companies think about their customer base?
Giacomo: The discussion with customers has changed in the past three to four years. Previously, it was about telling them that open source is secure because people are saying that they can find ways around the security.
But now, more people understand that open source software is probably more secure than non open source ones, because there are a lot of people that can fix things faster than if it was just done by one company.
While some CIOs still say they need to double-check the security of their controls, it’s now more about making sure that open source meets the needs of the lines of business that you’ve mentioned. Open source software is very good for IT infrastructure operators, but it has evolved more into helping developers push their code to production faster, for instance.
To simplify the open source landscape for customers, we provide an open source stack but customers can also use different modules from other stacks, such as other hypervisors besides KVM. The Suse Manager can manage Suse Linux and other Linux distributions, as well as Hyper-V or vSphere. We are trying to accommodate all that because in the real world, people run a variety of infrastructure components.
Open-source software suppliers are increasingly partnering with public cloud providers but at the same time, the cloud players are also providing proprietary services that are tied to their platforms. What can enterprises do to avoid this lock-in, which goes against the tenets of open source?
Giacomo: Open source can help enterprises avoid being locked into a stack in your datacentre, but now you can be locked into a public cloud – if you only use services from one cloud supplier, it can be very difficult to move out – and open source is trying to avoid that.
So, if you use containers with Kubernetes, you can move from Azure to AWS, Google Cloud or Alibaba Cloud or on-premise. You are not completely locked into one cloud service. If it’s getting more expensive at one cloud provider, you can move to another. And when it’s the opposite, you can go back.
But is it really that easy to switch cloud providers for container infrastructure, given that there are different Kubernetes distributions to choose from? Do most people use pure Kubernetes?
Giacomo: From a pure container infrastructure perspective, I would say yes. Most people use upstream Kubernetes. The difficulty comes when you start to use services around Kubernetes because they are very specific. So in public cloud number one, you will get specific AI services, databases or networking services on top of Kubernetes and those things are not very portable.
My recommendation would be to try to avoid consuming too many specific services and stay with open source. If you use PostgreSQL in one public cloud and Kubernetes, then you can move that to another public cloud without a problem. But if you start to use specific storage services from a public cloud, it will be more of a challenge.
In the storage space, we have a storage product based on the Ceph open source solution for block, object and file storage. It also has an AWS S3 storage interface so you can move your data from on-premise open source storage, to AWS and back on-premise when you need it. In some cases, you can still work around the lock-in by using open source bridges like that.
Thomas di Giacomo, Suse
We work a lot with cloud providers because our customers benefit from public cloud. In some use cases, bringing SAP Hana workloads to the public cloud makes sense, so we work things out with all the cloud suppliers.
On Azure, for instance, there are large instances of Hana which needs a lot of memory. So, we worked with Microsoft and SAP on Suse Linux for SAP Hana running on Azure, with up to 24 or 48 terabytes of memory instances.
Can you provide examples of use cases that are unique to Asia?
Giacomo: In Singapore, there’s a lot of innovation from the public sector, and a lot of investment in open source and the internet of things (IoT). It’s good to have open source innovation, but you need to build the platform. Users in Asia and Singapore don’t want to deal with PCs; they want to deal with solutions.
We also see a lot of use cases that combine AI and high performance computing (HPC), which Asia is very strong in. We’ve been customising Linux for HPC for a while now and we continue to do that.
What about running a lightweight Linux kernel at the edge of a network to do machine learning inferencing?
Giacomo: Suse Linux runs on different architectures including x86, IBM mainframe and ARM.
Two years ago, we released the same Linux for Raspberry Pi, which is a great device to do some of the IoT edge stuff like digital signage. The devices can compute and do some machine learning locally. If you are on the public cloud or run a datacentre, you can collect data from Raspberry Pi in different places and apply AI and analytics to improve business operations.
That sounds a lot like AWS’s Snowball Edge. Is Suse partnering with cloud providers that are also going into edge computing?
Giacomo: We play at the OS level, but we try to be agnostic and leave it open for customers to decide which cloud or hardware supplier they want to go with in their IoT implementations. There’s room for the industry to become more agnostic in the IoT space rather than push for siloed solutions.
For now, the IoT space is still quite new so that may not be a big problem, but in five to 10 years people will have a mix of IoT solutions from HPE, Dell and AWS which will all have to work together. That’s going to be interesting because I don’t think there’s a solution that can manage all of that today. I’m sure open source will help as well. In the Linux Foundation, there are IoT specific open source projects that are trying to address that challenge.
How will IBM’s acquisition of Red Hat affect Suse’s partnerships with Big Blue?
Giacomo: IBM has been a very strong partner of Suse for many years. It was the company that helped to put Linux in the enterprise through a $1bn investment in the late 90s. Suse and IBM were the first to put enterprise Linux on the IBM mainframe. We are still working very nicely with IBM today.
In the open source community, we work very closely with Red Hat. If you talk to my engineers working on the Linux kernel, whether it is Red Hat or Suse doesn’t really matter from a technical perspective. They are part of the same community – but to the sales people, that’s a different story.
We cannot predict what the future will be following the IBM-Red Hat deal, and if IBM will stay as independent as it was now that it can roll out its hardware, cloud and software together.
The other thing that sometimes comes into play is that we are not a US-based company. In Asia, especially in some parts of Asia, it’s actually interesting for Chinese companies to talk to us because we are not US headquartered.
How is Suse growing its customer base? Are they switching from another Linux vendor or going to Suse for specific use cases?
Giacomo: For generic workloads, Linux is pretty much the same across distributions – if you’re running a web server, it doesn’t really matter. For specific mission-critical workloads, such as SAP workloads, there are very clear differences between Suse Linux and others. We’ve been working with SAP for 20 years and we’ve optimised the technology, processes and installation management.
We are also very strong in the retail space, where we provide Suse Linux for point-of-sale terminals. We have Suse Manager dedicated to the retail industry, because retail shops sometimes don’t have good connectivity. If you need to manage thousands of shops and hundreds of cash machines running Linux, we have a dedicated Suse solution.
Besides the HPC use cases that I’ve mentioned, we also see a lot of interest in software-defined storage which is a lot cheaper than traditional storage. Open source software-defined storage is even cheaper than software-defined storage. At Suse, we don’t price open source software-defined storage by capacity. So unlike the other open source vendor offerings, the more data you need, the more cost-effective it is to use Suse.
OpenStack is another one. It’s not as cool as it used to be from an industry, marketing or even community standpoint, but we see growing adoption because the technology is mature. The last thing is improving the life of developers – and containers help a lot with that. However, developers still need help in developing applications and pushing out their code.
To that, we have Kubernetes that run containers, and we have Cloud Foundry that containerises itself and runs inside Kubernetes to provide developer tools on top of any Kubernetes distribution, whether it’s Azure Kubernetes Service (AKS), Suse Kubernetes or even OpenShift. That brings additional solutions to developers so that they can benefit from the container infrastructure as well.
Read more about open source
- Java Community Process chair Heather VanCura talks up efforts to keep Java up to speed with the needs of developers in the era of cloud and agile software development.
- Singapore telco and pay TV operator StarHub is the latest telco in the city-state to operate its network using OpenStack, underscoring the open source platform’s growing inroads in the telecoms industry.
- Red Hat CEO Jim Whitehurst claims Red Hat’s ability to support an entire technology stack is valuable to enterprises that want a common platform on-premise and in the cloud.
- Urs Hölzle, Google’s head of technical infrastructure, talks up efforts to deliver what he claims to be the first open source cloud stack.