Guoju Gao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Guoju Gao has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
- Yu-E Sun
Soochow University, Suzhou, China
25 shared publications - He Huang
Soochow University, Suzhou, China
22 shared publications - Yang Du
Soochow University, Suzhou, China
18 shared publications - Mingjun Xiao
School of Computer Science and Technology, University of Science and Technology, Hefei, China
14 shared publications - Jie Wu
Cloud Computing Research Institute, China Telecom
13 shared publications - Shigang Chen
University of Florida, Gainesville, USA
6 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 83-97
UniSketch: A Unified Sketch for Efficient General Flow Spread Measurement
- 14 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 83-97
Open AccessUniSketch: A Unified Sketch for Efficient General Flow Spread Measurement
- Open Access
- 18 May 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (3) , 1-26
Dynamic Flat Filter: A Unified Framework for Scalable and Stable Fingerprint-Based Filters
Fingerprint-based filters, such as Cuckoo and Quotient Filters, are a key class of data structures for approximate membership query with deletion support. However, their fixed capacity is a major limitation in dynamic environments with unpredictable data volumes. While prior research has introduced dynamic filters, their resizing mechanisms are tightly coupled to specific filter structures, making their mechanisms difficult to generalize. Moreover, prior dynamic solutions introduce severe performance bottlenecks: chain-based or ring-based filters enable incremental growth yet degrade queries by probing multiple sub-filters, while global-resizing filters cause long, disruptive stalls during resizing as their single, large filter grows. In addition, prior incremental designs often leave scale-induced false positive rate growth unaddressed. In this paper, we present the Dynamic Flat Filter (DFF), a unified framework that decouples dynamic resizing from core filter logic, enabling fingerprint-based filters to achieve incremental scalability while preserving their original performance, memory efficiency, and false positive guarantees. DFF maintains a variable-sized set of independent segments, each reusing the base filter's native layout, enabling fine-grained resizing with only minor integration hooks. We achieve O(1) insertion, query, and deletion by employing a flat structure for all segments and a lookup table that quickly maps any item to its corresponding segment. Additionally, DFF incorporates a fingerprint growth strategy that keeps the false positive rate essentially constant as the filter scales. We demonstrate the general applicability of DFF by integrating it with Cuckoo and Quotient Filters, each requiring fewer than 60 lines of code to be modified, and evaluate these implementations on real-world and synthetic datasets. Experimental results show that DFF outperforms SOTA baselines in insertion, query, and deletion, maintains high memory efficiency, and stabilizes the false positive rate. Notably, DFF attains at least 1.33× the overall throughput on query-intensive workloads and reduces worst-case insertion latency by up to 59.5% compared with the best prior dynamic filter.
- 18 May 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (3) , 1-26
Open AccessDynamic Flat Filter: A Unified Framework for Scalable and Stable Fingerprint-Based Filters
Fingerprint-based filters, such as Cuckoo and Quotient Filters, are a key class of data structures for approximate membership query with deletion support. However, their fixed capacity is a major limitation in dynamic environments with unpredictable data volumes. While prior research has introduced dynamic filters, their resizing mechanisms are tightly coupled to specific filter structures, making their mechanisms difficult to generalize. Moreover, prior dynamic solutions introduce severe performance bottlenecks: chain-based or ring-based filters enable incremental growth yet degrade queries by probing multiple sub-filters, while global-resizing filters cause long, disruptive stalls during resizing as their single, large filter grows. In addition, prior incremental designs often leave scale-induced false positive rate growth unaddressed. In this paper, we present the Dynamic Flat Filter (DFF), a unified framework that decouples dynamic resizing from core filter logic, enabling fingerprint-based filters to achieve incremental scalability while preserving their original performance, memory efficiency, and false positive guarantees. DFF maintains a variable-sized set of independent segments, each reusing the base filter's native layout, enabling fine-grained resizing with only minor integration hooks. We achieve O(1) insertion, query, and deletion by employing a flat structure for all segments and a lookup table that quickly maps any item to its corresponding segment. Additionally, DFF incorporates a fingerprint growth strategy that keeps the false positive rate essentially constant as the filter scales. We demonstrate the general applicability of DFF by integrating it with Cuckoo and Quotient Filters, each requiring fewer than 60 lines of code to be modified, and evaluate these implementations on real-world and synthetic datasets. Experimental results show that DFF outperforms SOTA baselines in insertion, query, and deletion, maintains high memory efficiency, and stabilizes the false positive rate. Notably, DFF attains at least 1.33× the overall throughput on query-intensive workloads and reduces worst-case insertion latency by up to 59.5% compared with the best prior dynamic filter.
- 7 March 2026
- Published by Elsevier in Computer Networks
- Vol. 280
MSFramework: Multi-stage similarity-based key flow identification in high-speed networks
- 7 March 2026
- Published by Elsevier in Computer Networks
- Vol. 280
MSFramework: Multi-stage similarity-based key flow identification in high-speed networks
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Networking
- Vol. 34, 7199-7214
SWVFM: Ranged Noise Removal for Sliding-Window Flow Cardinality Measurement in High-Speed Networks
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Networking
- Vol. 34, 7199-7214
SWVFM: Ranged Noise Removal for Sliding-Window Flow Cardinality Measurement in High-Speed Networks
- 15 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Big Data
- Vol. 12 (3) , 829-842
PieSketch: Finding Core Composition of Frequent Keys in High-Speed Data Streams
Key-value data streams are widely spread in modern information systems. This paper delves into the inner features of individual keys and introduces a new measurement task called Core Composition Estimation for Frequent Keys (CCEFK). CCEFK aims to identify the most significant values associated with an item’s key, which can be fully applied to a wide range of real-world scenarios. To address this quite challenging, multidimensional CCEFK mission, we propose PieSketch, a novel sketch-based algorithm that supports three types of queries related to CCEFK. PieSketch employs a three-level data structure and a two-staged insertion algorithm to process the keys and values of data streams separately and sequentially, achieving fine-grained data summarization for both. Through some rigorous theoretical analysis, we prove that PieSketch can achieve a good performance guarantee with low time and space complexity. We further present three optimizations to enhance PieSketch comprehensively: Sector Morphing addresses the diversity in value distribution among different keys with a dynamically adjusting strategy, significantly improving memory efficiency; Early Filtration ensures PieSketch’s stability in extreme conditions; SIMD Acceleration speeds up PieSketch for data processing. We finally conducted extensive experiments on four real-world datasets to evaluate PieSketch’s excellent performance. The results show that PieSketch significantly outperforms state-of-the-art techniques, achieving 23.67×, 88.41×, and 15.80× better accuracy on three types of queries, respectively, and demonstrating 1.84 times higher insertion throughput on average.
- 15 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Big Data
- Vol. 12 (3) , 829-842
PieSketch: Finding Core Composition of Frequent Keys in High-Speed Data Streams
Key-value data streams are widely spread in modern information systems. This paper delves into the inner features of individual keys and introduces a new measurement task called Core Composition Estimation for Frequent Keys (CCEFK). CCEFK aims to identify the most significant values associated with an item’s key, which can be fully applied to a wide range of real-world scenarios. To address this quite challenging, multidimensional CCEFK mission, we propose PieSketch, a novel sketch-based algorithm that supports three types of queries related to CCEFK. PieSketch employs a three-level data structure and a two-staged insertion algorithm to process the keys and values of data streams separately and sequentially, achieving fine-grained data summarization for both. Through some rigorous theoretical analysis, we prove that PieSketch can achieve a good performance guarantee with low time and space complexity. We further present three optimizations to enhance PieSketch comprehensively: Sector Morphing addresses the diversity in value distribution among different keys with a dynamically adjusting strategy, significantly improving memory efficiency; Early Filtration ensures PieSketch’s stability in extreme conditions; SIMD Acceleration speeds up PieSketch for data processing. We finally conducted extensive experiments on four real-world datasets to evaluate PieSketch’s excellent performance. The results show that PieSketch significantly outperforms state-of-the-art techniques, achieving 23.67×, 88.41×, and 15.80× better accuracy on three types of queries, respectively, and demonstrating 1.84 times higher insertion throughput on average.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 582-593
Memory-Efficient and Hardware-Friendly Sketches for Hierarchical Heavy Hitter Detection
Identifying the hierarchical heavy hitters (HHHs), i.e., the frequent aggregated flows based on common IP prefixes, is a vital task in network traffic measurement and security. Existing methods typically employ dynamic trie structures to track numerous prefixes or utilize multiple separate sketch instances, one for each hierarchical level, to capture HHHs across different levels, while both approaches suffer from low memory efficiency and limited compatibility with programmable switches. In this paper, we introduce two novel HHH detection solutions, respectively, Hierarchical Heavy Detector (HHD) and the Compressed Hierarchical Heavy Detector (CHHD), to achieve high memory efficiency and enhanced hardware compatibility. The key idea of HHD is to design a shared bucket array structure to identify and record HHHs from all hierarchical levels, which avoids the memory wastage of maintaining separate sketches to achieve high memory efficiency and allows feasible deployment of both byte-hierarchy and bit-hierarchy HHH detection on programmable switches using minimal processing stage resources. Additionally, HHD utilizes a sampling-based update strategy to effectively balance packet processing speed and detection accuracy. Furthermore, we present the CHHD, which enhances HHH detection in bit hierarchies through a more compact cell structure, which allows for compressing several ancestor and descendant prefixes within a single cell, further boosting memory efficiency and accuracy. We have implemented HHD and CHHD on a P4-based programmable switch with limited switch resources. Experimental results based on real-world Internet traces demonstrate that HHD and CHHD outperform the state-of-the-art by achieving up to 56 percentage points higher detection precision and 2.6× higher throughput.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 582-593
Memory-Efficient and Hardware-Friendly Sketches for Hierarchical Heavy Hitter Detection
Identifying the hierarchical heavy hitters (HHHs), i.e., the frequent aggregated flows based on common IP prefixes, is a vital task in network traffic measurement and security. Existing methods typically employ dynamic trie structures to track numerous prefixes or utilize multiple separate sketch instances, one for each hierarchical level, to capture HHHs across different levels, while both approaches suffer from low memory efficiency and limited compatibility with programmable switches. In this paper, we introduce two novel HHH detection solutions, respectively, Hierarchical Heavy Detector (HHD) and the Compressed Hierarchical Heavy Detector (CHHD), to achieve high memory efficiency and enhanced hardware compatibility. The key idea of HHD is to design a shared bucket array structure to identify and record HHHs from all hierarchical levels, which avoids the memory wastage of maintaining separate sketches to achieve high memory efficiency and allows feasible deployment of both byte-hierarchy and bit-hierarchy HHH detection on programmable switches using minimal processing stage resources. Additionally, HHD utilizes a sampling-based update strategy to effectively balance packet processing speed and detection accuracy. Furthermore, we present the CHHD, which enhances HHH detection in bit hierarchies through a more compact cell structure, which allows for compressing several ancestor and descendant prefixes within a single cell, further boosting memory efficiency and accuracy. We have implemented HHD and CHHD on a P4-based programmable switch with limited switch resources. Experimental results based on real-world Internet traces demonstrate that HHD and CHHD outperform the state-of-the-art by achieving up to 56 percentage points higher detection precision and 2.6× higher throughput.
- 20 October 2025
- Published by Springer Nature
- p. 359-378
FluxSketch: A Sketch-Based Solution for Long-Term Fluctuating Key Flow Detection
- 20 October 2025
- Published by Springer Nature
- p. 359-378
FluxSketch: A Sketch-Based Solution for Long-Term Fluctuating Key Flow Detection
- 9 September 2025
- Published by Oxford University Press (OUP) in The Computer Journal
- Vol. 69 (1) , 162-174
H2Sketch: real-time H -value measurement of key flows in high-speed networks
The identification of key flows, as one of the most important traffic measurement tasks in high-speed networks, has lots of practical applications, including network resource optimization and network attack detection. However, the existing works define the key flows only according to the flow size or the persistence (i.e. time dimension), all of which are one-sided definitions. To bridge the gap, we propose a new metric (named $H$-value) to identify key flows in high-speed networks, in which the flow $H$-value metric considers the joint distribution of flow size and time relationships. Specifically, the $H$-value of one flow means the flow size of this flow in at least $H$ measurement periods is not less than $H$. Based on this new metric, we redefine the key flows as the flows with large $H$-values, called $H^{2}$-flow. Efficiently and accurately identifying $H^{2}$-flows in high-speed networks faces significant challenges due to limited memory space, uneven distribution of flow $H$-values, and real-time requirements. To this end, we propose a new sketch structure called H$^{2}$Sketch, which consists of two main modules. Extensive experimental results based on real datasets show that H$^{2}$Sketch outperforms all the compared solutions in all evaluation metrics, such as ARE, F1 score, and throughput.- 9 September 2025
- Published by Oxford University Press (OUP) in The Computer Journal
- Vol. 69 (1) , 162-174
H2Sketch: real-time H -value measurement of key flows in high-speed networks
The identification of key flows, as one of the most important traffic measurement tasks in high-speed networks, has lots of practical applications, including network resource optimization and network attack detection. However, the existing works define the key flows only according to the flow size or the persistence (i.e. time dimension), all of which are one-sided definitions. To bridge the gap, we propose a new metric (named $H$-value) to identify key flows in high-speed networks, in which the flow $H$-value metric considers the joint distribution of flow size and time relationships. Specifically, the $H$-value of one flow means the flow size of this flow in at least $H$ measurement periods is not less than $H$. Based on this new metric, we redefine the key flows as the flows with large $H$-values, called $H^{2}$-flow. Efficiently and accurately identifying $H^{2}$-flows in high-speed networks faces significant challenges due to limited memory space, uneven distribution of flow $H$-values, and real-time requirements. To this end, we propose a new sketch structure called H$^{2}$Sketch, which consists of two main modules. Extensive experimental results based on real datasets show that H$^{2}$Sketch outperforms all the compared solutions in all evaluation metrics, such as ARE, F1 score, and throughput.- Open Access
- 30 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
Lightweight Two-level Collaborative Network Traffic Measurement for Data Center Networks
Network traffic measurement is crucial for the effective management of data center networks. Collaborative measurement solutions distribute measurement tasks to switches based on flow-level or packet-level granularity to alleviate the measurement load on each switch. However, flow-level solutions often experience severe imbalances in measurement overhead between switches measuring large or small flows, and face scalability challenges due to the costly optimization of collaborative plans. Additionally, packet-level solutions do not adequately reduce hash collisions in sketches and fail to significantly enhance measurement accuracy. In this paper, we present the Lightweight Two-level Collaborative Measurement (LTCM) that synergies flow-level and packet-level strategy to optimize measurement load balancing, reduce overall measurement overhead, and enhance measurement accuracy. We first design a Lightweight Flow-level Measurement (FCM) framework that balances the number of flows measured by each switch, incorporating a novel interval-matching technique that significantly lowers the computational costs of collaborative strategies. Based on FCM, LTCM implements our measurement load balancing strategy selector at ingress switches to detect flows whose number of packets entering the network exceeds a given threshold and evenly distribute their subsequent packets across all switches for measurement. To further improve measurement accuracy and speed, we design a two-layer collaborative sketch that reduces hash collisions between large and small flows. We implement LTCM on a Tofino-based programmable switch. Experimental results based on real Internet traces show that LTCM achieves highly efficient flow-level and packet-level measurement load balancing, improving accuracy by up to 96.6% and throughput by up to 112.79%. All related implementations are open-sourced.
- 30 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet Technology
Open AccessLightweight Two-level Collaborative Network Traffic Measurement for Data Center Networks
Network traffic measurement is crucial for the effective management of data center networks. Collaborative measurement solutions distribute measurement tasks to switches based on flow-level or packet-level granularity to alleviate the measurement load on each switch. However, flow-level solutions often experience severe imbalances in measurement overhead between switches measuring large or small flows, and face scalability challenges due to the costly optimization of collaborative plans. Additionally, packet-level solutions do not adequately reduce hash collisions in sketches and fail to significantly enhance measurement accuracy. In this paper, we present the Lightweight Two-level Collaborative Measurement (LTCM) that synergies flow-level and packet-level strategy to optimize measurement load balancing, reduce overall measurement overhead, and enhance measurement accuracy. We first design a Lightweight Flow-level Measurement (FCM) framework that balances the number of flows measured by each switch, incorporating a novel interval-matching technique that significantly lowers the computational costs of collaborative strategies. Based on FCM, LTCM implements our measurement load balancing strategy selector at ingress switches to detect flows whose number of packets entering the network exceeds a given threshold and evenly distribute their subsequent packets across all switches for measurement. To further improve measurement accuracy and speed, we design a two-layer collaborative sketch that reduces hash collisions between large and small flows. We implement LTCM on a Tofino-based programmable switch. Experimental results based on real Internet traces show that LTCM achieves highly efficient flow-level and packet-level measurement load balancing, improving accuracy by up to 96.6% and throughput by up to 112.79%. All related implementations are open-sourced.