arXiv subjects

Gil Einziger

Publications and source records attributed to Gil Einziger.

At least 19 recordsLinked to original sources

arXiv2025

Counter Pools: Counter Representation for Efficient Stream Processing

Due to the large data volume and number of distinct elements, space is often the bottleneck of many stream processing systems. The data structures used by these systems often consist of counters whose optimization yields significant memory savings. The challenge lies in balancing the size of the counters: too small, and they overflow; too large, and memory capacity limits their number. In this work, we suggest an efficient encoding scheme that sizes each counter according to its needs. Our approach uses fixed-sized pools of memory (e.g., a single memory word or 64 bits), where each pool manages a small number of counters. We pay special attention to performance and demonstrate considerable improvements for various streaming algorithms and workload characteristics.

cs.NI↗

arXiv2024

Floating-floating point: a highly accurate number representation with flexible Counting ranges

Efficient number representation is essential for federated learning, natural language processing, and network measurement solutions. Due to timing, area, and power constraints, such applications use narrow bit-width (e.g., 8-bit) number systems. The widely used floating-point systems exhibit a trade-off between the counting range and accuracy. This paper introduces Floating-Floating-Point (F2P) - a floating point number that varies the partition between mantissa and exponent. Such flexibility leads to a large counting range combined with improved accuracy over a selected sub-range. Our evaluation demonstrates that moving to F2P from the state-of-the-art improves network measurement accuracy and federated learning.

cs.NI↗

arXiv2024

SQUID: Faster Analytics via Sampled Quantile Estimation

Streaming algorithms are fundamental in the analysis of large and online datasets. A key component of many such analytic tasks is $q$-MAX, which finds the largest $q$ values in a number stream. Modern approaches attain a constant runtime by removing small items in bulk and retaining the largest $q$ items at all times. Yet, these approaches are bottlenecked by an expensive quantile calculation. This work introduces a quantile-sampling approach called SQUID and shows its benefits in multiple analytic tasks. Using this approach, we design a novel weighted heavy hitters data structure that is faster and more accurate than the existing alternatives. We also show SQUID's practicality for improving network-assisted caching systems with a hardware-based cache prototype that uses SQUID to implement the cache policy. The challenge here is that the switch's dataplane does not allow the general computation required to implement many cache policies, while its CPU is orders of magnitude slower. We overcome this issue by passing just SQUID's samples to the CPU, thus bridging this gap. In software implementations, we show that our method is up to 6.6x faster than the state-of-the-art alternatives when using real workloads. For switch-based caching, SQUID enables a wide spectrum of data-plane-based caching policies and achieves higher hit ratios than the state-of-the-art P4LRU.

cs.DS↗

arXiv2023

QUIC-FL: Quick Unbiased Compression for Federated Learning

Distributed Mean Estimation (DME), in which $n$ clients communicate vectors to a parameter server that estimates their average, is a fundamental building block in communication-efficient federated learning. In this paper, we improve on previous DME techniques that achieve the optimal $O(1/n)$ Normalized Mean Squared Error (NMSE) guarantee by asymptotically improving the complexity for either encoding or decoding (or both). To achieve this, we formalize the problem in a novel way that allows us to use off-the-shelf mathematical solvers to design the quantization.

cs.LG↗

arXiv2023

Uncertainty Estimation based on Geometric Separation

In machine learning, accurately predicting the probability that a specific input is correct is crucial for risk management. This process, known as uncertainty (or confidence) estimation, is particularly important in mission-critical applications such as autonomous driving. In this work, we put forward a novel geometric-based approach for improving uncertainty estimations in machine learning models. Our approach involves using the geometric distance of the current input from existing training inputs as a signal for estimating uncertainty, and then calibrating this signal using standard post-hoc techniques. We demonstrate that our method leads to more accurate uncertainty estimations than recently proposed approaches through extensive evaluation on a variety of datasets and models. Additionally, we optimize our approach so that it can be implemented on large datasets in near real-time applications, making it suitable for time-sensitive scenarios.

cs.LG↗

arXiv2022

Virtual Service Embedding with Time-Varying Load and Provable Guarantees

Deploying services efficiently while satisfying their quality requirements is a major challenge in network slicing. Effective solutions place instances of the services' virtual network functions (VNFs) at different locations of the cellular infrastructure and manage such instances by scaling them as needed. In this work, we address the above problem and the very relevant aspect of sub-slice reuse among different services. Further, unlike prior art, we account for the services' finite lifetime and time-varying traffic load. We identify two major sources of inefficiency in service management: (i) the overspending of computing resources due to traffic of multiple services with different latency requirements being processed by the same virtual machine (VM), and (ii) the poor packing of traffic processing requests in the same VM, leading to opening more VMs than necessary. To cope with the above issues, we devise an algorithm, called REShare, that can dynamically adapt to the system's operational conditions and find an optimal trade-off between the aforementioned opposite requirements. We prove that REShare has low algorithmic complexity and is asymptotic 2-competitive under a non-decreasing load. Numerical results, leveraging real-world scenarios, show that our solution outperforms alternatives, swiftly adapting to time-varying conditions and reducing service cost by over 25%.

cs.NI↗

arXiv2022

A Geometric Method for Improved Uncertainty Estimation in Real-time

Machine learning classifiers are probabilistic in nature, and thus inevitably involve uncertainty. Predicting the probability of a specific input to be correct is called uncertainty (or confidence) estimation and is crucial for risk management. Post-hoc model calibrations can improve models' uncertainty estimations without the need for retraining, and without changing the model. Our work puts forward a geometric-based approach for uncertainty estimation. Roughly speaking, we use the geometric distance of the current input from the existing training inputs as a signal for estimating uncertainty and then calibrate that signal (instead of the model's estimation) using standard post-hoc calibration techniques. We show that our method yields better uncertainty estimations than recently proposed approaches by extensively evaluating multiple datasets and models. In addition, we also demonstrate the possibility of performing our approach in near real-time applications. Our code is available at our Github https://github.com/NoSleepDeveloper/Geometric-Calibrator.

cs.LG↗

arXiv2022

On the Power of False Negative Awareness in Indicator-based Caching Systems

Distributed caching systems such as content distribution networks often advertise their content via lightweight approximate indicators (e.g., Bloom filters) to efficiently inform clients where each datum is likely cached. While false-positive indications are necessary and well understood, most existing works assume no false-negative indications. Our work illustrates practical scenarios where false-negatives are unavoidable and ignoring them has a significant impact on system performance. Specifically, we focus on false-negatives induced by indicator staleness, which arises whenever the system advertises the indicator only periodically, rather than immediately reporting every change in the cache. Such scenarios naturally occur, e.g., in bandwidth-constraint environments or when latency impedes the ability of each client to obtain an updated indicator. Our work introduces novel false-negative aware access policies that continuously estimate the false-negative ratio and sometimes access caches despite negative indications. We present optimal policies for homogeneous settings and provide approximation guarantees for our algorithms in heterogeneous environments. We further perform an extensive simulation study with multiple real system traces. We show that our false-negative aware algorithms incur a significantly lower access cost than existing approaches or match the cost of these approaches while requiring an order of magnitude fewer resources (e.g., caching capacity or bandwidth).

cs.NI↗

arXiv2022

False Negative Awareness in Indicator-based Caching Systems

Distributed caching systems such as content distribution networks often advertise their content via lightweight approximate indicators (e.g., Bloom filters) to efficiently inform clients where each datum is likely cached. While false-positive indications are necessary and well understood, most existing works assume no false-negative indications. Our work illustrates practical scenarios where false-negatives are unavoidable and ignoring them significantly impacts system performance. Specifically, we focus on false-negatives induced by indicator staleness, which arises whenever the system advertises the indicator only periodically, rather than immediately reporting every change in the cache. Such scenarios naturally occur, e.g., in bandwidth-constraint environments or when latency impedes each client's ability to obtain an updated indicator. Our work introduces novel false-negative aware access policies that continuously estimate the false-negative ratio and sometimes access caches despite negative indications. We present optimal policies for homogeneous settings and provide approximation guarantees for our algorithms in heterogeneous environments. We further perform an extensive simulation study with multiple real system traces. We show that our false-negative aware algorithms incur a significantly lower access cost than existing approaches or match the cost of these approaches while requiring an order of magnitude fewer resources (e.g., caching capacity or bandwidth).

cs.NI↗

arXiv2022

Parallel Virtual Machines Placement with Provable Guarantees

Network Function Virtualization (NFV) carries the potential for on-demand deployment of network algorithms in virtual machines (VMs). In large clouds, however, VM resource allocation incurs delays that hinder the dynamic scaling of such NFV deployment. Parallel resource management is a promising direction for boosting performance, but it may significantly increase the communication overhead and the decline ratio of deployment attempts. Our work analyzes the performance of various placement algorithms and provides empirical evidence that state-of-the-art parallel resource management dramatically increases the decline ratio of deterministic algorithms but hardly affects randomized algorithms. We, therefore, introduce APSR -- an efficient parallel random resource management algorithm that requires information only from a small number of hosts and dynamically adjusts the degree of parallelism to provide provable decline ratio guarantees. We formally analyze APSR, evaluate it on real workloads, and integrate it into the popular OpenStack cloud management platform. Our evaluation shows that APSR matches the throughput provided by other parallel schedulers, while achieving up to 13x lower decline ratio and a reduction of over 85% in communication overheads.

cs.NI↗

arXiv2022

Self-Adjusting Ego-Trees Topology for Reconfigurable Datacenter Networks

State-of-the-art topologies for datacenters (DC) and high-performance computing (HPC) networks are demand-oblivious and static. Therefore, such network topologies are optimized for the worst-case traffic scenarios and can't take advantage of changing demand patterns when such exist. However, recent optical switching technologies enable the concept of dynamically reconfiguring circuit-switched topologies in real-time. This capability opens the door for the design of self-adjusting networks: networks with demand-aware and dynamic topologies in which links between nodes can be established and re-adjusted online and respond to evolving traffic patterns. This paper studies a recently proposed model for optical leaf-spine reconfigurable networks. We present a novel algorithm, GreedyEgoTrees, that dynamically changes the network topology. The algorithm greedily builds ego trees for nodes in the network, where nodes cooperate to help each other, taking into account the global needs of the network. We show that GreedyEgoTrees has nice theoretical properties, outperforms other possible algorithms (like static expander and greedy dynamic matching) and can significantly improve the average path length for real DC and HPC traces.

cs.NI↗

arXiv2021

Limited Associativity Makes Concurrent Software Caches a Breeze

Software caches optimize the performance of diverse storage systems, databases and other software systems. Existing works on software caches automatically resort to fully associative cache designs. Our work shows that limited associativity caches are a promising direction for concurrent software caches. Specifically, we demonstrate that limited associativity enables simple yet efficient realizations of multiple cache management schemes that can be trivially parallelized. We show that the obtained hit ratio is usually similar to fully associative caches of the same management policy, but the throughput is improved by up to X5 compared to production-grade caching libraries, especially in multi-threaded executions.

cs.AR↗

arXiv2021

Lightweight Robust Size Aware Cache Management

Modern key-value stores, object stores, Internet proxy caches, as well as Content Delivery Networks (CDN) often manage objects of diverse sizes, e.g., blobs, video files of different lengths, images with varying resolution, and small documents. In such workloads, size-aware cache policies outperform size-oblivious algorithms. Unfortunately, existing size-aware algorithms tend to be overly complicated and computationally~expensive. Our work follows a more approachable pattern; we extend the prevalent (size-oblivious) TinyLFU cache admission policy to handle variable sized items. Implementing our approach inside two popular caching libraries only requires minor changes. We show that our algorithms yield competitive or better hit-ratios and byte hit-ratios compared to the state of the art size-aware algorithms such as AdaptSize, LHD, LRB, and GDSF. Further, a runtime comparison indicates that our implementation is faster by up to x3 compared to the best alternative, i.e., it imposes much lower CPU overhead.

cs.OS↗

arXiv2021

Self-adjusting Advertisement of Cache Indicators with Bandwidth Constraints

Cache advertisements reduce the access cost by allowing users to skip the cache when it does not contain their datum. Such advertisements are used in multiple networked domains such as 5G networks, wide area networks, and information-centric networking. The selection of an advertisement strategy exposes a trade-off between the access cost and bandwidth consumption. Still, existing works mostly apply a trial-and-error approach for selecting the best strategy, as the rigorous foundations required for optimizing such decisions is lacking. Our work shows that the desired advertisement policy depends on numerous parameters such as the cache policy, the workload, the cache size, and the available bandwidth. In particular, we show that there is no ideal single configuration. Therefore, we design an adaptive, self-adjusting algorithm that periodically selects an advertisement policy. Our algorithm does not require any prior information about the cache policy, cache size, or workload, and does not require any apriori configuration. Through extensive simulations, using several state-of-the-art cache policies, and real workloads, we show that our approach attains a similar cost to that of the best static configuration (which is only identified in retrospect) in each case.

cs.NI↗

arXiv2021

SALSA: Self-Adjusting Lean Streaming Analytics

Counters are the fundamental building block of many data sketching schemes, which hash items to a small number of counters and account for collisions to provide good approximations for frequencies and other measures. Most existing methods rely on fixed-size counters, which may be wasteful in terms of space, as counters must be large enough to eliminate any risk of overflow. Instead, some solutions use small, fixed-size counters that may overflow into secondary structures. This paper takes a different approach. We propose a simple and general method called SALSA for dynamic re-sizing of counters and show its effectiveness. SALSA starts with small counters, and overflowing counters simply merge with their neighbors. SALSA can thereby allow more counters for a given space, expanding them as necessary to represent large numbers. Our evaluation demonstrates that, at the cost of a small overhead for its merging logic, SALSA significantly improves the accuracy of popular schemes (such as Count-Min Sketch and Count Sketch) over a variety of tasks. Our code is released as open-source [1].

cs.DS↗

arXiv2020

Routing Oblivious Measurement Analytics

Network-wide traffic analytics are often needed for various network monitoring tasks. These measurements are often performed by collecting samples at network switches, which are then sent to the controller for aggregation. However, performing such analytics without ``overcounting'' flows or packets that traverse multiple measurement switches is challenging. Therefore, existing solutions often simplify the problem by making assumptions on the routing or measurement switch placement. We introduce AROMA, a measurement infrastructure that generates a uniform sample of packets and flows regardless of the topology, workload and routing. Therefore, AROMA can be deployed in many settings, and can also work in the data plane using programmable PISA switches. The AROMA infrastructure includes controller algorithms that approximate a variety of essential measurement tasks while providing formal accuracy guarantees. Using extensive simulations on real-world network traces, we show that our algorithms are competitively accurate compared to the best existing solutions despite the fact that they make no assumptions on the underlying network or the placement of measurement switches.

cs.NI↗

arXiv2020

Faster and More Accurate Measurement through Additive-Error Counters

Counters are a fundamental building block for networking applications such as load balancing, traffic engineering, and intrusion detection, which require estimating flow sizes and identifying heavy hitter flows. Existing works suggest replacing counters with shorter multiplicative error \emph{estimators} that improve the accuracy by fitting more of them within a given space. However, such estimators impose a computational overhead that degrades the measurement throughput. Instead, we propose \emph{additive} error estimators, which are simpler, faster, and more accurate when used for network measurement. Our solution is rigorously analyzed and empirically evaluated against several other measurement algorithms on real Internet traces. For a given error target, we improve the speed of the uncompressed solutions by $5\times$-$30\times$, and the space by up to $4\times$. Compared with existing state-of-the-art estimators, our solution is $ 9\times$-$35\times$ faster while being considerably more accurate.

cs.DS↗

arXiv2019

Reducing Service Deployment Cost Through VNF Sharing

Thanks to its computational and forwarding capabilities, the mobile network infrastructure can support several third-party ("vertical") services, each composed of a graph of virtual (network) functions (VNFs). Importantly, one or more VNFs are often common to multiple services, thus the services deployment cost could be reduced by letting the services share the same VNF instance instead of devoting a separate instance to each service. By doing that, however, it is critical that the target KPI (key performance indicators) of all services are met. To this end, we study the VNF sharing problem and make decisions on (i) when sharing VNFs among multiple services is possible, (ii) how to adapt the virtual machines running the shared VNFs to the combined load of the assigned services, and (iii) how to prioritize the services traffic within shared VNFs. All decisions aim to minimize the cost for the mobile operator, subject to requirements on end-to-end service performance, e.g., total delay. Notably, we show that the aforementioned priorities should be managed dynamically and vary across VNFs. We then propose the FlexShare algorithm to provide near-optimal VNF-sharing and priority assignment decisions in polynomial time. We prove that FlexShare is within a constant factor from the optimum and, using real-world VNF graphs, we show that it consistently outperforms baseline solutions.

cs.NI↗