Systems Research@Google (SRG)

Shaping the future of hyperscaler systems design.

SRG Mission

The SystemsResearch@Google (SRG) team works in the heart of Google's AI and Infrastructure organization. Our mission is to help shape the future of hyperscaler systems design for Google by inventing, incubating, and infusing new concepts, designs, and technologies into Google applications, systems, and data centers. Tight integration with engineering and product teams ensures a multidisciplinary approach to problem solving, and we also maintain strong relationships with external research communities working on pressing systems-research problems.

We drive fundamental advances in software systems, hardware architecture, distributed systems, systems for AI and machine learning, ML/AI for systems, databases and analytics, networking, security, privacy, memory and storage systems, and sustainability.

Extreme memory efficiency

Improving memory efficiency by making memory tiering more effective

Architectural features for future processors

Identify & explore innovative architectural options

Hyperscale OS structures

Rethinking OS and distributed system designs to improve data center efficiency and security

ML and generative AI for databases & analytics

Reimagining data management systems leveraging AI

Efficient systems for agent learning and serving

Unlocking ML, RL & Agentic efficiencies through new frameworks

Agent security and privacy

Ensure confidentiality, safety, and integrity of AI agents

Transforming existing software infrastructure to Rust

AI and agents for transformation to Rust in systems software

Next-generation storage systems

Enhanced efficiency for hyperscale and AI storage

“If you work in computer systems, it can be difficult to fully understand cloud computing unless you see it from the inside. The real-world requirements of global scale, extreme reliability, fault tolerance, security and privacy, plus the exponential growth of data and demand for machine learning ... that’s really unique and amazing to see up close.”

Hank Levy

Distinguished Engineer, SRG Director • Google AI and Infrastructure

Featured research

IEEE MICRO Sept 2026 (Micro Top Picks)

Concorde: Fast and Accurate CPU Performance Modeling with Compositional Analytical-ML Fusion

Arash Nasr-Esfahany, Mohammad Alizadeh, Victor Lee, Hanna Alam, Brett W. Coon, David Culler, Vidushi Dadu, Martin Dixon, Henry M. Levy, Santosh Pandey, Parthasarathy Ranganathan, Amir Yazdanbakhsh.

View details

OSDI 26 (Best Paper Honorable Mention)

MDK: Rethinking the Data Center Memory Reclamation Problem

Shaurya Patel, Suli Yang, Yawen Wang, Kan Wu, Alexandra Fedorova, Margo Seltzer, Kimberly Keeton.

View details

PACMMOD 26

100x Cost & Latency Reduction: Performance Analysis of AI Query Approximation Using Lightweight Proxy Models

Y. Chung, R. Desai, J. He, Y. Xiao, T. Hottelier, Y.-L. Samo, P. Khadilkar, X. Chen, S. Idicula, F. Ozcan, A. Halevy, Y. Papakonstantinou

View details

OSDI 26

OBASE: Object-Based Address Space Engineering to Improve Memory Tiering

Vinay Banakar, Suli Yang, Kan Wu, Andrea Arpaci-Dusseau, Remzi Arpaci-Dusseau, Kimberly Keeton.

View details
SIGMOD 26

An In-Depth Study of Filter-Agnostic Vector Search on a PostgreSQL Database System

Duo Lu, Helena Caminal, Manos Chatzakis, Yannis Papakonstantinou, Yannis Chronis, Vaibhav Jain, Fatma Ozcan.

View details
NSDI 26

Crosscheck: Input Validation for WAN Control Systems

Alexander Krentsel, Rishabh Iyer, Isaac Keslassy, Sylvia Ratnasamy, Bharath Modhipalli, Anees Shaikh, and Rob Shakir.

View details
RLBoost

Harvesting Preemptible Cloud Resources for Cost-Efficient Reinforcement Learning on LLMs

Yongji Wu, Xueshen Liu, Haizhong Zheng, Juncheng Gu, Beidi Chen, Z. Morley Mao, Arvind Krishnamurthy, Ion Stoica.

View details

Publications

IEEE MICRO: Concorde: Fast and Accurate CPU Performance Modeling with Compositional Analytical-ML Fusion.

Arash Nasr-Esfahany, Mohammad Alizadeh, Victor Lee, Hanna Alam, Brett W. Coon, David Culler, Vidushi Dadu, Martin Dixon, Henry M. Levy, Santosh Pandey, Parthasarathy Ranganathan, Amir Yazdanbakhsh.

OSDI 26: MDK: Rethinking the Data Center Memory Reclamation Problem.

Shaurya Patel, Suli Yang, Yawen Wang, Kan Wu, Alexandra Fedorova, Margo Seltzer, Kimberly Keeton.

OSDI 26: OBASE: Object-Based Address Space Engineering to Improve Memory Tiering.

Vinay Banakar, Suli Yang, Kan Wu, Andrea Arpaci-Dusseau, Remzi Arpaci-Dusseau, Kimberly Keeton.

OSDI 26: Scaling the IO Wall with Declarative IO.

Sanjith Athlur, Sara McAllister, Theo Gregersen, Timothy Kim, Yiwei Chen, Sarvesh Tandon, Lucy Wang, Daniel S. Berger, Saurabh Kadekodi, Arif Merchant, Benjamin Berg, Nathan Beckmann, Rashmi Vinayak, George Amvrosiadis, Gregory R. Ganger.

PACMMOD 26: 100x Cost & Latency Reduction: Performance Analysis of AI Query Approximation Using Lightweight Proxy Models.

Y. Chung, R. Desai, J. He, Y. Xiao, T. Hottelier, Y.-L. Samo, P. Khadilkar, X. Chen, S. Idicula, F. Ozcan, A. Halevy, Y. Papakonstantinou.

SIGMOD 26: An In-Depth Study of Filter-Agnostic Vector Search on a PostgreSQL Database System.

Duo Lu, Helena Caminal, Manos Chatzakis, Yannis Papakonstantinou, Yannis Chronis, Vaibhav Jain, Fatma Ozcan.

NSDI 26: Crosscheck: Input Validation for WAN Control Systems.

Alexander Krentsel, Rishabh Iyer, Isaac Keslassy, Sylvia Ratnasamy, Bharath Modhipalli, Anees Shaikh, and Rob Shakir.

NSDI 26: RLBoost: Harvesting Preemptible Cloud Resources for Cost-Efficient Reinforcement Learning on LLMs.

Yongji Wu, Xueshen Liu, Haizhong Zheng, Juncheng Gu, Beidi Chen, Z. Morley Mao, Arvind Krishnamurthy, Ion Stoica.

Eurosys 26: TCO-driven Storage Provisioning for Exascale Data Centers.

Timothy Kim, Saurabh Kadekodi, Arif Merchant, Prashant Nema, Jai Menon, K. V. Rashmi, Gregory R. Ganger.

NINeS 26: No Signal to Rule Them All: A Systematic Analysis of In-Network Congestion Signals. Sarah McClure, Nandita Dukkipati, Sylvia Ratnasamy, Scott Shenker.

Sarah McClure, Nandita Dukkipati, Sylvia Ratnasamy, Scott Shenker.

ASPLOS 26: Wave: Offloading Resource Management to SmartNIC Cores. J. Humphries, N. Natu, K. Kaffes, S. Novakovic, P. Turner, H. Levy, D. Culler, C. Kozyrakis.

J. Humphries, N. Natu, K.Kaffes, S. Novakovic, P. Turner, H. Levy, D. Culler, C. Kozyrakis.

CIDR 26: Declarative Memory Services.

Jeronimo Castillon, Jana Giceva, Yu Hua, Kimberly Keeton, Akhil Shekar, Kevin Skadron, Tianzheng Wang, Huanchen Zhang.

SOSP 25: IC-Cache: Efficient Large Language Model Serving via In-context Caching.

Y. Yu, Y. Gan, L. Tsai, N. Sarda, J. Shen, Y. Zhou, A. Krishnamurthy, F. Lai, H. Levy, D. Culler.

VLDB 25: Is Long Context All You Need? Leveraging LLM's Extended Context for NL2SQL. Yeounoh Chung, Gaurav Kakkar, Yu Gan, Brenton Milne, Fatma Özcan.

Yeounoh Chung, Gaurav Kakkar, Yu Gan, Brenton Milne, Fatma Özcan,

VLDB 25: Filtered Vector Search: State-of-the-Art and Research Opportunities.

Yannis Chronis, Helena Caminal, Yannis Papakonstantinou, Fatma Özcan, and Anastasia Ailamaki.

USENIX ATC 25: PageFlex: Flexible and Efficient User-space Delegation of Linux Paging Policies with eBPF.

Anil Yelam, Kan Wu, Zhiyuan Guo, Suli Yang, Rajath Shashidhara, Wei Xu, Stanko Novaković, Alex C. Snoeren, Kimberly Keeton.

ISCA 25: Concorde: Fast and Accurate CPU Performance Modeling with Compositional Analytical-ML Fusion.

Arash Nasr-Esfahany, Mohammad Alizadeh, Victor Lee, Hanna Alam, Brett W. Coon, David Culler, Vidushi Dadu, Martin Dixon, Henry M. Levy, Santosh Pandey, Parthasarathy Ranganathan, Amir Yazdanbakhsh.

ISCA 25: RAGO: Systematic Performance Optimization for Retrieval-Augmented Generation Serving.

W. Jiang, S. Subramanian, C. Graves, G. Alonso, A. Yazdanbakhsh, and V. Dadu.

HotOS 25: Contextual Agent Security: A Policy for Every Purpose.

Lily Tsai, Eugene Bagdasarian.

MLSys 25: LAVA: Lifetime-Aware VM Allocation with Learned Distributions and Adaptation to Mispredictions.

J. Ling, P. Worah, Y. Wang, Y. Kong, C. Wang, C. Stein, D. Gupta, J. Behmer, L. Bush, P. Ramanaman, R. Kumar, T. Chestna, Y. Liu, Y. Zhao, K. McKinley, M. Park, M. Mass.

ICLR 25: CHASE-SQL: Multi-path Reasoning and Preference Optimized Candidate Selection in Text-to-SQL.

M. Pourreza, H. Li, R. Sun, Y. Chung, S. Talaei, G. T. Kakkar, Y. Gan, A. Saberi, F. Özcan, Sercan Arik.

IEEE Data Eng.: LLMs and Databases: A Synergistic Approach to Data Utilization.

F. Özcan, Y. Cheung, Y. Chronis, Y. Gan, Y. Wang, C. Binnig, J. Wehrstein, G. T. Kakkar, and S. Abu-el-haija.

CIDR 25: Towards Foundation Database Models.

Johannes Wehrstein, Carsten Binnig, Fatma Özcan, Shobha Vasudevan, Yu Gan, Yawen Wang.

SIGCOMM 24: A Decentralized SDN Architecture for the WAN.

A. Krentsel, N. Saran, B. Koley, S Mandal, A.Narayanan, S. Ratnasamy, A. Al-Shabibi, A. Shaikh, R. Shakir, A. Singla, H. Weatherspoon.

ASPLOS 24: Sharing is leaking: blocking transient-execution attacks with core-gapped confidential VMs.

Charly Castes and Andrew Baumann.

ASPLOS 24: CC-NIC: a Cache-Coherent Interface to the NIC.

Henry Schuh, Arvind Krishnamurthy, David Culler, Henry Levy, Luigi Rizzo, Samira Khan, Brent Stephens.

ASPLOS 24: Limoncello: Prefetchers for Scale.

Akanksha Jain, Hannah Lin, Carlos Villavieja, Baris Kasikci, Chris Kennelly, Milad Hashemi, Parthasarathy Ranganathan.

ASPLOS 23: Towards an Adaptable Systems Architecture for Memory Tiering at Warehouse Scale.

Padmapriya Duraisamy, Wei Xu, Scott Hare, Ravi Rajwar, David Culler, Zhiyi Xu, Jianing Fan, Chris Kennelly, Bill McCloskey, Danijela Mijalovic, Brian Morris, Chiranjit Mukherjee, Jiangliang Ren, Greg Thalen, Paul Turner, Carlos Villavieja, Parthasarathy Ranganathan, Amin Vahdat.

NSDI 23: Invisinets: Removing Networking from Cloud Networks.

Sarah McClure, Zeke Medley, Deepak Bansal, Karthick Jayaraman, Ashok Narayanan, Jitendra Padhye, Sylvia Ratnasamy, Anees Shaikh, RIshabh Tewari.

ISCA 23: Profiling Hyperscale Big Data Processing.

Abraham Gonzalez, Aasheesh Kolli, Samira Khan, Sihang Liu, Vidushi Dadu, Sagar Karandikar, Jichuan Chang, Krste Asanovic, Parthasarathy Ranganathan.

SOSP 23: A Cloud-Scale Characterization of Remote Procedure Calls.

Korakit Seemakhupt, Brent Stephens, Samira Khan, Sihang Liu, Hassan Wassel, Soheil Yeganeh, Alex Snoeren, Arvind Krishnamurthy, David Culler, Henry Levy.

VLDB 23: DINOMO: An Elastic, Scalable, High-Performance Key-Value Store for Disaggregated Persistent Memory.

Sekwon Lee, Soujanya Ponnapalli, Sharad Singhal, Marcos Aguilera, Kimberly Keeton, Vijay Chidambaram.

VLDB 23: WiscSort: External Sorting For Byte-Addressable Storage.

Vinay Banakar, Kan Wu, Yuvraj Patel, Kimberly Keeton, Andrea C Arpaci-Dusseau, Remzi H Arpaci-Dusseau.

VLDB Journal 23: HERMES: data placement and schema optimization for enterprise knowledge bases.

Chuan Lei, Abdul Quamar, Vasilis Eftymiou, Fatma Özcan, Rana Alotaibi.

HotNets 23: A Case Against CXL Memory Pooling.

Philip Levis, Kun Lin, Amy Tai.

OSDI 22: Carbink: Fault-Tolerant Far Memory.

Yang Zhou, Hassan Wassel, Sihang Liu, Jiaqi Gao, James Mickens, Minlan Yu, Chris Kennelly, Paul Turner, David Culler, Henry Levy, Amin Vahdat.

HotNets 22: Understanding Host Interconnect Congestion.

Aksham Agarwal, Rachit Agarwal, Behnam Montazeri, Masoud Moshref, Khaled Elmeleegy, Luigi Rizzo, Marc Asher de Kruijf, Gautam Kumar, Sylvia Ratnasamy, David Culler, Amin Vahdat.

NVMW 22: SoftPM: Software Persistent Memory.

Yuanchao Xu, Wei Xu, Kimberly Keeton, David Culler.

VLDB 22: Reflections on my data management research journey (VLDB Women in Database Research Award Talk).

Fatma Özcan.

HotOS 21: Cores that don’t count.

Peter Hochschild, Paul Turner, Jeffrey Mogul, Rama Govindaraju, Parthasarathy Ranganathan, David Culler, Amin Vahdat.

“SRG is perfectly situated to help drive change in collaboration with the great engineers at Google creating AI and infrastructure systems at unprecedented scale. It is an exciting time to be a systems researcher, with new designs and approaches needed in response to unprecedented demands, resource shortages, and exciting AI advances. AI also gives us powerful new tools for systems research and design.”