In an exclusive interview with NCN Magazine, Mr. Sudhir Tangri, Vice President & General Manager – Asia Pacific Sales, Keysight Technologies, discusses the latest data center trends in India, role of ML & AI in data centers, cyber-attack types data centers face, the cybersecurity solutions offered, and how their solutions are future-proof against evolving threats.
What are the latest trends while building data centers with high availability in the Indian context?
As the data center industry continues to experience significant growth in India, several key trends are emerging around managing power consumption, increasing capacity and optimizing utilization. Government initiatives, corporate investment, increased startup support and a greater focus on collaboration with educational institutions are all driving growth in this area. However, this growth also means it is crucial to address resource challenges. Techniques like data parallelism, tensor, and pipeline parallelism as well as implementation and optimization of collective operations algorithms help in improving data center utilization. This in turn is also placing a greater focus on the benchmarking of data center hosts and the rigorous testing of network fabric.
How are AI and ML changing data center management?
Data center management involves efficient functioning and operations of AI host infrastructure, AI network fabric, and associated software and services to ensure 24×7 uptime, maximum capacity utilization, minimal power consumption, and security and compliance. AI data centers are of a much larger scale due to the intensive computing requirements of ML algorithms and AI models. Key performance indicators for these data centers include Job Completion Time (JCT) while reducing power consumption. With this in mind, it is essential to look for tools that can determine JCT in various AI worlds.
How big is the market for Edge Data Centers in India?
The operation of AI / ML network infrastructure has two important components. This is made up of backend data center for AI model training and front-end data center for inference workloads, also referred to as edge data centers. Edge data centers generally process north-south network traffic and have ultra-low latency requirements with high-speed access and secure data in motion. Considering the level of digitalization and proliferation of UPI-based financial services, education, and healthcare needs, India has the vast potential for edge data centers. Government, BFSI, and other services companies are implementing edge data centers to provide effective and timely services.
What are the future trends in the data center sector?
The quest for AI dominance is a significant global endeavor. In addition, India is poised for a leap in data center infrastructure due to the massive digitalization of services. However, the challenges in the design, deployment and operations of these high scale data centers need to be addressed as we evolve. Research has already started for 1.6, 3.2THz direct and coherent interfaces, PCIe7 and CXL 3.0 as well as DDR7 to cater to high-speed interconnect for AI data centers. Keysight is passionate about helping customers in this journey and is already testing these high-speed interfaces with our comprehensive test portfolio consisting of BERT system for up to 120 GBd (NRZ & PAM4), 110GHz UXR Scope and 256 GS/s & 80 GHz BW AWG. Keysight also offers traffic and protocol testing on 800GE on AresONE platform.
Most recently, Keysight has introduced Keysight AI (KAI) Data Center Builder, an advanced software suite that emulates real-world workloads to evaluate how new algorithms, components, and protocols impact the performance of AI training. KAI Data Center Builder’s workload emulation capability integrates large language model (LLM) and other AI model training workloads into the design and validation of AI infrastructure components — networks, hosts, and accelerators. This solution enables tighter synergy between hardware design, protocols, architectures, and AI training algorithms, boosting system performance.
What are the most common cyber-attacks that data centers encounter and what specific cybersecurity solutions do you offer for protection? In addition to the traditional cyber-attacks such as malware, ransomware, DDoS, phishing, social engineering attacks and insider threats, AI infrastructure is vulnerable to AI specific attacks like data poisoning, evasion attacks, model inversion and adversarial attacks. These kinds of attacks pose significant challenges to AI data centers. Therefore, the security assessment of any AI data center needs to be a continuous activity to ascertain security posture against these kinds of novel and challenging attacks. Keysight offers a wide range of cyber security test solutions including Threat Simulator, BreakPoint, and CyPerf which can help in determining the security posture of AI data centers.
Covered By: NCN MAGAZINE / Keysight
If you have an interesting Article / Report/case study to share, please get in touch with us at editors@roymediative.com , roy@roymediative.com, 9811346846/ 9625243429
