Developing the most relevant solutions for its customers is of utmost importance to IBM’s Linux Technology Center in India. This requires constant interaction and collaboration with them, in order to improve continuously and innovate constantly. In an exclusive chat, Dipankar Sarma, distinguished engineer, Linux Technology Center – IBM India, talks about the Center’s role with respect to Linux and its impact on IBM Power Systems, among other things, with Rahul Chopra, editorial director, EFY Group.
Q. Could you share the overall profile of your team that is present in India and globally?
A. I lead Linux development for IBM Power Systems. I am the technical leader for a worldwide team of developers who work on the Linux platform. So, we work with the Linux kernel, its distributions, and various other open source capabilities that go with Linux and also run on our IBM Power Systems portfolio. That portfolio includes a number of different types of servers such as high-end servers, and large systems that run enterprise applications like SAP HANA. These are at the high end of the spectrum. But the portfolio goes all the way down to the smaller servers which we run for AI and cloud workloads. These systems are meant for a wide range of applications.
We have had a presence in India from the time our organisation was established 20 years ago. We have a very mature team with lots of contributions coming in from the worldwide team as well as from the Indian team. We continue to bring out new innovations for the platforms that run on Linux and lots of new optimisations for it.
Q. Is your team solely focused on developing the product, or does it also have interactions with clients and puts in an effort to support them?
A. As an organisation, we are focused on the product development side related to systems development, and specifically, Linux development. For example, we make the necessary changes in the Linux kernel. And then through our Linux distribution partners such as Red Hat, those products reach our customers.
While working on the product R&D, we ensure that the features our customers have asked for are developed and delivered. For this, we go out and spend time talking to them on how they are using our systems, and about what difficulties they have to face. Our customer interactions include working with various CIOs, CTOs and customer councils. We also spend a lot of time interacting with IT managers and IT architects, who are involved in implementing and maintaining the infrastructures for such systems. Their feedback helps us in improving our systems and providing better products.
Q. Does this mean that the Linux leadership for the IBM Power Systems platform is being driven from India?
A. I would say that a lot of contributions come from the worldwide team spread across multiple geographies such as the US, India, Australia, Brazil, France, and so on. The 2017 Linux Kernel Development Report ranks us (IBM) among the top five in the world, in terms of contributions to the Linux kernel. We have consistently been among the top five contributors for many years, and have sustained this by working on the kernel with our global team of Linux engineers across several geographies.
Q. What are the key open source contributions made by your team at IBM?
A. A considerable number of contributions have been made to the Linux kernel. There are many people working on Linux on the two platforms, IBM Power Systems and IBM Z. The contributions range from core Linux kernel (scheduler/mm/file systems) to key architecture enablements for Power Systems and Z (mainframe) hardware.
It is to be noted here that our open source contributions are not just limited to Linux. Our hybrid cloud portfolio (IBM Cloud and IBM Cloud Private – the company’s flagship cloud portfolio) is entirely based on open source. There are a lot of IBM contributors to open source outside the Linux team – making contributions to blockchain frameworks such as Hyperledger, container cloud orchestration frameworks such as Kubernetes, and others supported by the Linux Foundation.
Some of the interesting contributions that have been made in recent times have been around AI, particularly AI Fairness 360 (AIF360), which brings in a certain amount of fairness by checking for unwanted bias and mitigating it in machine learning.
An important toolkit called ART, or Adversarial Robustness Toolbox, allows you to protect your developments against adversarial attacks using bad data. SystemML is another big example of a machine learning framework that has received heavy contributions from those at IBM.
Many contributions have been made by us, along with our collaborators, to IBM’s AI Cloud Services (also known as Watson Machine Learning), which is based on open source frameworks. So quite a lot of open source contributions are being made from across the business units of IBM.
Another quick mention is about the quantum computing toolkit (Qiskit), which allows you to build quantum programs and manipulate circuits.
“As an organisation, we are focused on the product development side related to systems development”
Q. Is the access to the quantum computing infrastructure available in India directly, or is it connected to some other region?
A. At the moment, access in India is available on the cloud. The actual IBM quantum systems are not available in India. They are available in the cloud for academic institutions and other partners, so that they can test and experiment with them.
Q. Is there a possibility or any initiative towards customising Linux for quantum computing? Or will a whole different set of software need to be developed for it?
A. I think this space will evolve, and Linux won’t be required to run for quantum computing in the future. However, traditional systems will be required to work hand-in-hand with the quantum systems. Currently, traditional systems are going to continue to innovate the way they are at the moment.
Q. Has an organisation like IBM built a culture of motivating its teams to contribute, or is it just natural for developers to contribute?
A. I would say that it’s a bit of both. We encourage people with a natural inclination towards contribution to collaborate and work with open source communities. Often, people want to work on open source projects, for which we give them that opportunity.
Over the years, we have built a culture of working in open source communities and a strategy of giving back to these communities, with respect to the technologies that we work upon.
We have a long history of doing this and continue to do so, as it is very important to us. As a part of this culture, we encourage people to participate and present their open source contributions. That is why it is important for our teams to go out to open source conferences and interact with people.
With 20 plus years of history, this culture is now kind of ingrained in our working style and helps us to become good citizens in the open source community.
Q. While hiring, do you look at whether people have contributed to an open source project? And is this factor also considered for appraisals?
A. It’s not necessary that one should have contributed to open source. But if such a contribution has been made, then it is definitely a plus point while hiring. And when we do look at people’s open source contribution experience, then we focus on what are their contributions that will help them work with the community.
Q. Do you prefer to hire freshers or people with a certain level of experience for your team?
A. We usually hire across the spectrum, which includes fresh engineering graduates and experienced people.
Q. In your opinion, how evolved is India with respect to the open source and Linux ecosystem? Do we have the required talent or is that one of the challenges you face as an engineering leader?
A. When we started out 20 years ago, there were very few contributors from India in the Linux kernel community. There used to be occasional contributions from some of the IITs.
But since then, the contributions have increased a lot. We see a considerable amount of exposure to Linux at the university level, especially in Tier 1 and Tier 2 engineering colleges. This familiarity with Linux has enabled people to do projects and take up credit courses based on Linux.
It is also worth mentioning that we see very good participation from people, especially from the technology cities of India. So, I think the trend is looking good. But there’s always more that can be done with respect to training more and more people at the university level.
Overall, my observation is that in India, it has been a very good journey for Linux over the last 15 to 20 years.
Q. Is it correct to assume that now the discussion about Linux and open source, with a CIO or CTO, is much easier as compared to what used to be the case earlier?
A. Indeed, that’s true. According to a survey on open source conducted by Red Hat this year, 89 per cent of the enterprises surveyed consider that open source plays an important role in their businesses. And 67 per cent have adopted open source for non-cost reasons.
For them, open source is bringing about lots of new innovations in a very short span of time due to its agility, security, robustness and reliability.
According to this survey, 59 per cent of the enterprises believe that the use of open source will increase in their organisations. So this is a good indication of the trend of using open source in enterprises, and is getting reflected in CIO and CTO interactions too.
Q. How is Linux growing in terms of its share on the Power Systems platform with the introduction of SAP HANA kind of applications?
A. Linux has been there for quite some time (approximately 20 years) on the power systems. Back then it was running on the IBM Hypervisor named PowerVM in a virtualised environment. In 2014, Linux on IBM Power Systems spread out from its high-performance computing domain to other areas with a focus on the cloud, AI and enterprise databases like SAP HANA. And that trend is sort of continuing with the growth of our business.
The same holds true with respect to the mainframe, System Z, which is running Linux as well. A lot of incredible technologies have been brought onto the System Z platform for scaling to hundreds of billions of transactions and providing end-to-end security. So, considerable progress has been made on both these platforms in terms of adopting and optimising Linux. Compared to the Linux market’s growth, IBM’s Linux business is growing at a faster rate.
Q. Within your IBM systems (both Z and Power Systems platforms), is the share of Linux servers increasing vis-a-vis your other solutions?
A. On a year-to-year basis, our Linux business is growing considerably well at the moment. This growth has been a part of our long-term strategy. Regarding the IBM Power series of servers running Linux, we have had a presence in the HPC market in the past where a lot of top performing HPC systems have been IBM power systems. About six years ago, a number of our partners, like Mellanox, Google and Rackspace from the OpenPOWER Foundation, started collaborating with us on the possibility that these systems could also be very good for the cloud and AI market. And on discovering new markets and new possibilities related to AI and the cloud, we directed our efforts on these new technologies and built a successful product line.
Q. According to you, what are the key technology trends that drive the evolution of Linux and keep it relevant for the future?
A. Linux’s advantage has been that it’s so configurable that it can be run on pretty much any kind of platform, all the way from embedded systems to very large enterprise-class systems.
We’re seeing Linux actually being used not just in the main system, but also running on many other parts of a server system, like the embedded controllers that manage a server system and I/O adapters.
It is well known by now that reliability and high-availability technologies in Linux are very useful for high-end enterprise software and solutions. But there is also work going on to make Linux a very good player for the cloud.
Memory technologies are a good example of new innovations coming to Linux. Memory is no longer limited to just the DRAM technology. Within the system memory, we now have different classes of memory technologies, e.g., DRAM, storage class memory, etc — all of which have different performance characteristics and different semantic behaviour. So, a lot of work is happening in Linux right now, to handle multiple memory technologies in a single system.
So, I see this trend of a lot of new Linux innovations emerging that are based on new system technologies, be it on the processor architecture side or the memory architecture side.
“Many contributions have been made by us, along with our collaborators, to IBM’s AI Cloud Services (also known as Watson Machine Learning), which is based on open source frameworks”.
Q. Could you tell us a bit more about the OpenPOWER Foundation?
A. The OpenPOWER Foundation is a global organisation comprising like-minded industry and academic partners. It includes people who work on hardware, chip design, system design and system engineering, all the way up to cloud software, high-performance computing software and academic research. Currently, it has over 300 members.
The OpenPOWER Foundation began as a small organisation initiated by partners like IBM, Mellanox, NVIDIA and Google. The organisation slowly expanded to include a large pool of hardware and software vendors. Some of the partners, like Suzhou PowerCore from China, belong to the Asia-Pacific region.
Q. Does the organisation operate in India?
A. OpenPOWER does have academic partners from several universities in India, who are working on research projects around the OpenPOWER platform. For example, we have a research collaboration with one of the OpenPOWER Foundation academic members in India, who is from the National Institute of Technology, Surathkal, Karnataka. They are working on the popular open source architecture simulator called gem5, and adding power processor and system architecture support in it. For an academic research group to have such a modern architecture simulator, which can do performance evaluations, is very important. The OpenPOWER Foundation is quite supportive of this work as well.
