Introduction:
Vinod Khosla, a prominent figure in the high-tech industry, delivered a keynote speech at ONS (Optical Networking Summit). Known for his contributions to groundbreaking companies like Sun Microsystems, NextGen, Juniper, and Serent, Khosla shared his insights on disruptive innovations, challenges incumbent giants, and the potential impact of his current ventures.

Sun Microsystems:
Co-founder of Sun Microsystems, Khosla introduced merchant silicon open source to workstations. This bold move challenged the dominance of computer vendors in the early 1980s.

NextGen:
Supported and funded NextGen, a company that combined the best of CISC and RISC processors. This venture challenged Intel’s dominance in the processor market.

Juniper:
Incubated and funded Juniper, which brought ASIC-based solutions to IP routers. Juniper’s innovation challenged Cisco in the core router business, creating a three-peat of disruptive successes.

Serent:
Supported Serent in the late 1990s, transforming the transport infrastructure dominated by legacy telecom vendors.

Current Ventures:
Emphasizing the potential impact of Khosla’s current ventures in clean tech, the speaker highlighted the potential for these bets to have a significant impact on society.

Challenges in Networking:
Khosla noted the lack of significant changes in networking since the late 1990s, suggesting SDN (Software-Defined Networking) as a potential game-changer.

Conclusion:
The presentation concluded with anticipation surrounding Khosla’s upcoming remarks, highlighting his expertise in disruptive innovations and the potential impact of his current ventures.

Abstractions and the Evolution of Engineering:
Vinod Khosla emphasizes the importance of abstraction in understanding complex systems. He draws parallels between the challenges faced in networking today and those in databases 15 years ago. Khosla argues that engineering methodologies need to adapt to prioritize change, evolvability, and experimentation.

The Cathedral vs. the Bazaar:
Khosla highlights the need for organizations to embrace change and experimentation to remain competitive. He stresses that avoiding risks can ultimately lead to taking the biggest risk of all.

Engineering for Change:
Khosla proposes that the focus of engineering should shift from optimizing for cost, performance, or reliability to optimizing for change. This requires engineering methodologies to evolve and prioritize evolvability, specialization, and experimentation.

Modularity and Failure Resilience:
Modularity is essential for building systems that can be easily modified and updated. Khosla emphasizes the importance of designing systems to fail gracefully, gradually, and in small steps, rather than attempting to prevent failures altogether.

Complexity in Configuration and Provisioning:
Khosla highlights the staggering complexity involved in configuration and provisioning of networking systems, requiring up to 80,000 lines of code.

Security in the New Computing Environment:
Khosla stresses the need to re-engineer security approaches in the context of SDN and the new computing environment. He suggests considering endpoint security, encompassing mobile devices and storage blocks, as a potential solution.

Reaching the Limits of Human Understanding:
Khosla warns that as complexity increases, we are reaching the limits of human understanding. This can limit the time, scale, and performance achievable in complex systems.

SDN Solutions:
Vinod Khosla gathered input from several experts on the problems with Software-Defined Networks (SDNs) and solutions to address them. While preserving legacy systems is essential, inventing for the future is crucial. Specific solutions are less important than re-engineering engineering methodologies to adapt to SDN.

SDX Capabilities:
SDN encompasses software-defined storage, computing, and virtualization as its foundation. SDN enables a range of capabilities, such as network function virtualization, which are more feasible with SDN.

The Importance of SDN:
SDN provides a foundation for provably correct software, a critical aspect given the increasing importance of networks in our lives. It establishes a stronger intellectual foundation, defines appropriate abstractions, and enables systematic bug identification and tracking.

Re-engineering Engineering:
The main focus of Khosla’s talk is re-engineering engineering practices. Unpredictable workloads and the rise of big data and VDI demand greater evolvability rather than optimizing for specific current workloads.

Innovation and Experimentation:
As technology becomes more integrated into our lives, the number and diversity of devices accessing networks will increase. Experimentation is essential for innovation, but it requires reducing the risk of system failures.

Optimization Limits:
Cisco’s optimization focus led to complex systems with bugs, hindering further optimizations.

Obsolescence Planning:
Engineers need to create obsolescence strategies for systems, including dismantling methods.

Flexibility Prioritization:
Engineers should prioritize flexibility over performance when the rate of change is high.

Pac-Man Budget:
The CIO’s budget resembles Pac-Man, where operating costs consume innovation.

Autonomic Systems:
Machine learning systems should manage IT, network management, and other tasks.

Complexity and Hackability:
Complexity in software like Windows, the financial system, and video games increases their vulnerability to hacks.

Modularity and Abstraction in Computing Systems:
Complexity in computing introduces vulnerabilities to hacking, demanding innovative security solutions. Modularity, with small modules, is a potential direction to mitigate these risks, but there’s no definitive answer at present.

Google’s Approach to Data Center Management:
Google’s decision to develop its own Software Defined Networks (SDNs) and Software Defined Exchange (SDX) was crucial in experimenting and maintaining system reliability. Google App Engine goes beyond Amazon’s level of abstraction, leading to a discussion on the need for a data center operating system.

The Future of Computing:
Computing is becoming increasingly diverse, with millions of applications emerging. Lookout, a leader in mobile security, observes billions of unique application versions monthly, highlighting the vastness of the app landscape.

SDN: A Catalyst for Innovation:
SDN is the beginning of the software era, enabling innovation and simplifying complexity. It allows for disaggregation and abstraction of computing, storage, and networking components, viewing them as latencies in the system.

Fractal Systems and Data Analytics:
The concept of fractal systems, where each part resembles the whole, influences the future of computing. Real-time data analytics engines, such as Splunk, provide deep workload characterization and insights. Distributed analytics engines, like in-memory distributed databases, can process vast data volumes efficiently.

Evolutionary Systems and Openness:
Clay Sharkey’s notion of open systems emphasizes their evolvability, enabling adaptation to changing requirements. Evolutions assume that partial implementations have utility and that change is inevitable, rendering initial designs obsolete.

Linux as a Model of Evolutionary Development:
Linux’s development resembles a bazaar, contrasting with the cathedral model of systematic development. Linux’s success illustrates the power of the bazaar model in the face of change, highlighting the need for re-engineering engineering processes.

Modular Development and Peer Review:
Modular development, successive refinement, and aggressive peer review are essential for rapid evolution, facilitated by an appropriate architecture. Periodic architecture changes are necessary to overcome complexity barriers.

Founding Juniper Networks:
In 1996, Vinod Khosla, co-founder of Sun Microsystems, invested in Netscape and Amazon, recognizing the potential of the internet. Despite skepticism from carriers and Cisco, Khosla saw the need for a TCPIP-based internet and founded Juniper Networks. Juniper’s routers introduced evolvability and modularity by separating the control plane from the forwarding plane.

Open Systems vs Traditional Wisdom:
Khosla emphasizes the importance of challenging conventional wisdom and embracing innovation. He credits Pradeep Sindhu’s decision to reinvent TCPIP routing as a key factor in Juniper’s success. Khosla stresses the significance of adaptability, agility, and momentum, which are enabled by modularity and abstraction.

SDN and a New Era of Software and Networking:
Khosla highlights the narrow focus of software-defined networking (SDN) and the need for broader abstractions. He recommends Scott Schenker’s talk on a new era of software and networking, emphasizing its relevance to ongoing industry developments.

Accelerating Rate of Change and Entrepreneurship:
Khosla predicts an acceleration in the rate of change, creating opportunities for innovation and entrepreneurship. He draws a distinction between incremental change and transformative change, likening it to the difference between a wind and a typhoon. Khosla concludes by expressing gratitude and inviting questions from the audience.

Research versus Ventures:
Attendees may expect venture capitalists to focus on vision, but Vinod Khosla emphasizes the importance of translating research into practical applications.

Evolvable Systems:
Khosla recommends readings on evolvable systems by Clay Sharkey to provide insights beyond research-level knowledge.

Importance of Use:
In contrast to research, putting a system into use forces it to evolve and adapt based on user feedback and requirements.

Open Source Methodology:
Khosla sees open source methodology as a precursor to re-engineering engineering by facilitating rapid evolution through real-world usage.

Small Market Opportunities:
He highlights that small niche markets with exponential growth potential can be more interesting than large markets with more established competition.

Tolerating Missing Features:
In niche markets, users may tolerate missing features due to critical needs, allowing products to grow into larger markets.

Cisco’s iOS as an Example:
Apple’s iOS is seen as innovative, while Cisco’s iOS is perceived as outdated, underscoring the significance of evolution in engineering.

“Circus Midget and Fossilized Dinosaur Turd” Article:
Khosla references this article, emphasizing the need for engineering and system evolution to avoid becoming irrelevant.

Large Organizations and Risk Management:
Large organizations prioritize risk management and focus on mitigating risks that could impact revenue or service level agreements (SLAs). Moving the mainstream of an organization to adopt new technologies carries significant risk and can be challenging.

Importance of Experimentation:
Encouraging experimentation is crucial for innovation. Experimentation should be conducted in small niches where the consequences of failure are minimal while the potential for success is substantial.

Centralized vs. Distributed Systems:
Evolvability and rapid change are critical factors in IT systems. Distributed systems tend to be more reliable and adaptable compared to centralized systems. Modularity is essential for building adaptable and resilient systems.

OpenFlow and SDN:
OpenFlow and SDN are components of a broader approach to creating modular, self-healing, and autonomic systems. The goal is to enable IT administrators to manage more systems effectively using distributed and modular approaches.

Control and Automation:
Centralized control may not be necessary in modern IT systems. Autonomic and self-healing systems can adapt and respond to changes without direct human intervention. The control console may serve primarily as a monitoring tool, with human intervention required only for exceptional situations.

Evolvability and Modularity:
Modularity and evolvability are critical for future-proofing IT systems. Systems should be designed to accommodate rapid changes and evolving requirements.