arXiv subjects

Michael Mitzenmacher

Publications and source records attributed to Michael Mitzenmacher.

At least 37 records · Page 2Linked to original sources

arXiv2025

HACK: Homomorphic Acceleration via Compression of the Key-Value Cache for Disaggregated LLM Inference

Disaggregated Large Language Model (LLM) inference has gained popularity as it separates the computation-intensive prefill stage from the memory-intensive decode stage, avoiding the prefill-decode interference and improving resource utilization. However, transmitting Key-Value (KV) data between the two stages can be a bottleneck, especially for long prompts. Additionally, the computation time overhead for prefill and decode is key for optimizing Job Completion Time (JCT), and KV data size can become prohibitive for long prompts and sequences. Existing KV quantization methods can alleviate the transmission bottleneck and reduce memory requirements, but they introduce significant dequantization overhead, exacerbating the computation time. We propose Homomorphic Acceleration via Compression of the KV cache (HACK) for disaggregated LLM inference. HACK eliminates the heavy KV dequantization step, and directly performs computations on quantized KV data to approximate and reduce the cost of the expensive matrix-multiplication step. Extensive trace-driven experiments show that HACK reduces JCT by up to 70.9% compared to disaggregated LLM inference baseline and by up to 52.3% compared to state-of-the-art KV quantization methods.

cs.DC↗

arXiv2025

Efficient $d$-ary Cuckoo Hashing at High Load Factors by Bubbling Up

A $d$-ary cuckoo hash table is an open-addressed hash table that stores each key $x$ in one of $d$ random positions $h_1(x), h_2(x), \ldots, h_d(x)$. In the offline setting, where all items are given and keys need only be matched to locations, it is possible to support a load factor of $1 - ε$ while using $d = \lceil \ln ε^{-1} + o(1) \rceil$ hashes. The online setting, where keys are moved as new keys arrive sequentially, has the additional challenge of the time to insert new keys, and it has not been known whether one can use $d = O(\ln ε^{-1})$ hashes to support $\poly(ε^{-1})$ expected-time insertions. In this paper, we introduce bubble-up cuckoo hashing, an implementation of $d$-ary cuckoo hashing that achieves all of the following properties simultaneously: (1) uses $d = \lceil \ln ε^{-1} + α\rceil$ hash locations per item for an arbitrarily small positive constant $α$. (2) achieves expected insertion time $O(δ^{-1})$ for any insertion taking place at load factor $1 - δ\le 1 - ε$. (3) achieves expected positive query time $O(1)$, independent of $d$ and $ε$. The first two properties give an essentially optimal value of $d$ without compromising insertion time. The third property is interesting even in the offline setting: it says that, even though \emph{negative} queries must take time $d$, positive queries can actually be implemented in $O(1)$ expected time, even when $d$ is large.

cs.DS↗

arXiv2024

FRANCIS: Fast Reaction Algorithms for Network Coordination In Switches

Optimizing the reaction to network events, which is critical in tasks such as clock synchronization, multicast, and routing, becomes increasingly challenging as networks grow larger. To improve the reaction time compared to centralized solutions, the theory community has made significant progress in the design of message-passing algorithms that leverage all nodes for distributed computation, and the advent of programmable switches makes it now possible to materialize them. We propose FRANCIS, a framework and associated libraries for running message-passing algorithms on programmable switches. It features primitives that allow easy integration of such algorithms for quickly reacting to network events while optimizing resource consumption. We use FRANCIS to implement event reaction solutions that improve clock synchronization, source-routed multicast, and routing and demonstrate up to 18x reduction in reaction time.

cs.NI↗

arXiv2024

Beyond Throughput and Compression Ratios: Towards High End-to-end Utility of Gradient Compression

Gradient aggregation has long been identified as a major bottleneck in today's large-scale distributed machine learning training systems. One promising solution to mitigate such bottlenecks is gradient compression, directly reducing communicated gradient data volume. However, in practice, many gradient compression schemes do not achieve acceleration of the training process while also preserving accuracy. In this work, we identify common issues in previous gradient compression systems and evaluation methodologies. These include excessive computational overheads; incompatibility with all-reduce; and insufficient evaluation methods, such as not using an end-to-end metric or using a 32-bit baseline instead of the stronger 16-bit baseline. We revisit common compression approaches (sparsification, quantization, and low-rank decomposition) and demonstrate how considering the above issues can lead to minor but strategic design changes, resulting in notably better performance. Our goal is to raise awareness of the need for design and evaluation standards that naturally translate to the end-to-end utility of gradient compression.

cs.LG↗

arXiv2024

Fast Inference for Augmented Large Language Models

Augmented Large Language Models (LLMs) enhance the capabilities of standalone LLMs by integrating external data sources through API calls. In interactive LLM applications, efficient scheduling is crucial for maintaining low request completion times, directly impacting user engagement. However, these augmentations introduce scheduling challenges due to the need to manage limited memory for cached information (KV caches). As a result, traditional size-based scheduling algorithms, such as Shortest Job First (SJF), become less effective at minimizing completion times. Existing work focuses only on handling requests during API calls by preserving, discarding, or swapping memory without considering how to schedule requests with API calls. In this paper, we propose LAMPS, a novel LLM inference framework for augmented LLMs. LAMPS minimizes request completion time through a unified scheduling approach that considers the total length of requests and their handling strategies during API calls. Recognizing that LLM inference is memory-bound, our approach ranks requests based on their consumption of memory over time, which depends on both the output sizes and how a request is managed during its API calls. To implement our scheduling, LAMPS predicts the strategy that minimizes memory waste of a request during its API calls, aligning with but improving upon existing approaches. We also propose starvation prevention techniques and optimizations to mitigate the overhead of our scheduling. We implement LAMPS on top of vLLM and evaluate its performance against baseline LLM inference systems, demonstrating improvements in end-to-end latency by 27%-85% and reductions in TTFT by 4%-96% compared to the existing augmented-LLM system, with even greater gains over vLLM.

cs.LG↗

arXiv2024

Don't Stop Me Now: Embedding Based Scheduling for LLMs

Efficient scheduling is crucial for interactive Large Language Model (LLM) applications, where low request completion time directly impacts user engagement. Size-based scheduling algorithms like Shortest Remaining Process Time (SRPT) aim to reduce average request completion time by leveraging known or estimated request sizes and allowing preemption by incoming jobs with shorter service times. However, two main challenges arise when applying size-based scheduling to LLM systems. First, accurately predicting output lengths from prompts is challenging and often resource-intensive, making it impractical for many systems. As a result, the state-of-the-art LLM systems default to first-come, first-served scheduling, which can lead to head-of-line blocking and reduced system efficiency. Second, preemption introduces extra memory overhead to LLM systems as they must maintain intermediate states for unfinished (preempted) requests. In this paper, we propose TRAIL, a method to obtain output predictions from the target LLM itself. After generating each output token, we recycle the embedding of its internal structure as input for a lightweight classifier that predicts the remaining length for each running request. Using these predictions, we propose a prediction-based SRPT variant with limited preemption designed to account for memory overhead in LLM systems. This variant allows preemption early in request execution when memory consumption is low but restricts preemption as requests approach completion to optimize resource utilization. On the theoretical side, we derive a closed-form formula for this SRPT variant in an M/G/1 queue model, which demonstrates its potential value. In our system, we implement this preemption policy alongside our embedding-based prediction method.

cs.LG↗

arXiv2024

Learning-Augmented Frequency Estimation in Sliding Windows

We show how to utilize machine learning approaches to improve sliding window algorithms for approximate frequency estimation problems, under the ``algorithms with predictions'' framework. In this dynamic environment, previous learning-augmented algorithms are less effective, since properties in sliding window resolution can differ significantly from the properties of the entire stream. Our focus is on the benefits of predicting and filtering out items with large next arrival times -- that is, there is a large gap until their next appearance -- from the stream, which we show improves the memory-accuracy tradeoffs significantly. We provide theorems that provide insight into how and by how much our technique can improve the sliding window algorithm, as well as experimental results using real-world data sets. Our work demonstrates that predictors can be useful in the challenging sliding window setting.

cs.DS↗

arXiv2024

Learning-Based Heavy Hitters and Flow Frequency Estimation in Streams

Identifying heavy hitters and estimating the frequencies of flows are fundamental tasks in various network domains. Existing approaches to this challenge can broadly be categorized into two groups, hashing-based and competing-counter-based. The Count-Min sketch is a standard example of a hashing-based algorithm, and the Space Saving algorithm is an example of a competing-counter algorithm. Recent works have explored the use of machine learning to enhance algorithms for frequency estimation problems, under the algorithms with prediction framework. However, these works have focused solely on the hashing-based approach, which may not be best for identifying heavy hitters. In this paper, we present the first learned competing-counter-based algorithm, called LSS, for identifying heavy hitters, top k, and flow frequency estimation that utilizes the well-known Space Saving algorithm. We provide theoretical insights into how and to what extent our approach can improve upon Space Saving, backed by experimental results on both synthetic and real-world datasets. Our evaluation demonstrates that LSS can enhance the accuracy and efficiency of Space Saving in identifying heavy hitters, top k, and estimating flow frequencies.

cs.DS↗

arXiv2024

Tabula: Efficiently Computing Nonlinear Activation Functions for Secure Neural Network Inference

Multiparty computation approaches to secure neural network inference commonly rely on garbled circuits for securely executing nonlinear activation functions. However, garbled circuits require excessive communication between server and client, impose significant storage overheads, and incur large runtime penalties. To reduce these costs, we propose an alternative to garbled circuits: Tabula, an algorithm based on secure lookup tables. Our approach precomputes lookup tables during an offline phase that contains the result of all possible nonlinear function calls. Because these tables incur exponential storage costs in the number of operands and the precision of the input values, we use quantization to reduce these storage costs to make this approach practical. This enables an online phase where securely computing the result of a nonlinear function requires just a single round of communication, with communication cost equal to twice the number of bits of the input to the nonlinear function. In practice our approach costs 2 bytes of communication per nonlinear function call in the online phase. Compared to garbled circuits with 8-bit quantized inputs, when computing individual nonlinear functions during the online phase, experiments show Tabula with 8-bit activations uses between $280$-$560 \times$ less communication, is over $100\times$ faster, and uses a comparable (within a factor of 2) amount of storage; compared against other state-of-the-art protocols Tabula achieves greater than $40\times$ communication reduction. This leads to significant performance gains over garbled circuits with quantized inputs during the online phase of secure inference of neural networks: Tabula reduces end-to-end inference communication by up to $9 \times$ and achieves an end-to-end inference speedup of up to $50 \times$, while imposing comparable storage and offline preprocessing costs.

cs.CR↗

arXiv2024

THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression

Deep neural networks (DNNs) are the de facto standard for essential use cases, such as image classification, computer vision, and natural language processing. As DNNs and datasets get larger, they require distributed training on increasingly larger clusters. A main bottleneck is the resulting communication overhead where workers exchange model updates (i.e., gradients) on a per-round basis. To address this bottleneck and accelerate training, a widely-deployed approach is compression. However, previous deployments often apply bi-directional compression schemes by simply using a uni-directional gradient compression scheme in each direction. This results in significant computational overheads at the parameter server and increased compression error, leading to longer training and lower accuracy. We introduce Tensor Homomorphic Compression (THC), a novel bi-directional compression framework that enables the direct aggregation of compressed values and thus eliminating the aforementioned computational overheads. Moreover, THC is compatible with in-network aggregation (INA), which allows for further acceleration. Our evaluation shows that training representative vision and language models with THC reaches target accuracy by 1.40x to 1.47x faster using INA and 1.28x to 1.33x faster using a software PS compared with state-of-the-art systems.

cs.LG↗

arXiv2024

SkipPredict: When to Invest in Predictions for Scheduling

In light of recent work on scheduling with predicted job sizes, we consider the effect of the cost of predictions in queueing systems, removing the assumption in prior research that predictions are external to the system's resources and/or cost-free. In particular, we introduce a novel approach to utilizing predictions, SkipPredict, designed to address their inherent cost. Rather than uniformly applying predictions to all jobs, we propose a tailored approach that categorizes jobs based on their prediction requirements. To achieve this, we employ one-bit "cheap predictions" to classify jobs as either short or long. SkipPredict prioritizes predicted short jobs over long jobs, and for the latter, SkipPredict applies a second round of more detailed "expensive predictions" to approximate Shortest Remaining Processing Time for these jobs. Our analysis takes into account the cost of prediction. We examine the effect of this cost for two distinct models. In the external cost model, predictions are generated by some external method without impacting job service times but incur a cost. In the server time cost model, predictions themselves require server processing time, and are scheduled on the same server as the jobs.

cs.LG↗

arXiv2023

Analyzing Generalized Pólya Urn Models using Martingales, with an Application to Viral Evolution

The randomized play-the-winner (RPW) model is a generalized Pólya Urn process with broad applications ranging from clinical trials to molecular evolution. We derive an exact expression for the variance of the RPW model by transforming the Pólya Urn process into a martingale, correcting an earlier result of Matthews and Rosenberger (1997). We then use this result to approximate the full probability mass function of the RPW model for certain parameter values relevant to genetic applications. Finally, we fit our model to genomic sequencing data of SARS-CoV-2, demonstrating a novel method of estimating the viral mutation rate that delivers comparable results to existing scientific literature.

stat.AP↗

arXiv2023

Direct Telemetry Access

Fine-grained network telemetry is becoming a modern datacenter standard and is the basis of essential applications such as congestion control, load balancing, and advanced troubleshooting. As network size increases and telemetry gets more fine-grained, there is a tremendous growth in the amount of data needed to be reported from switches to collectors to enable network-wide view. As a consequence, it is progressively hard to scale data collection systems. We introduce Direct Telemetry Access (DTA), a solution optimized for aggregating and moving hundreds of millions of reports per second from switches into queryable data structures in collectors' memory. DTA is lightweight and it is able to greatly reduce overheads at collectors. DTA is built on top of RDMA, and we propose novel and expressive reporting primitives to allow easy integration with existing state-of-the-art telemetry mechanisms such as INT or Marple. We show that DTA significantly improves telemetry collection rates. For example, when used with INT, it can collect and aggregate over 400M reports per second with a single server, improving over the Atomic MultiLog by up to $16$x.

cs.NI↗

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↗

arXiv2022

Proteus: A Self-Designing Range Filter

We introduce Proteus, a novel self-designing approximate range filter, which configures itself based on sampled data in order to optimize its false positive rate (FPR) for a given space requirement. Proteus unifies the probabilistic and deterministic design spaces of state-of-the-art range filters to achieve robust performance across a larger variety of use cases. At the core of Proteus lies our Contextual Prefix FPR (CPFPR) model - a formal framework for the FPR of prefix-based filters across their design spaces. We empirically demonstrate the accuracy of our model and Proteus' ability to optimize over both synthetic workloads and real-world datasets. We further evaluate Proteus in RocksDB and show that it is able to improve end-to-end performance by as much as 5.3x over more brittle state-of-the-art methods such as SuRF and Rosetta. Our experiments also indicate that the cost of modeling is not significant compared to the end-to-end performance gains and that Proteus is robust to workload shifts.

cs.DB↗

arXiv2022

EDEN: Communication-Efficient and Robust Distributed Mean Estimation for Federated Learning

Distributed Mean Estimation (DME) is a central building block in federated learning, where clients send local gradients to a parameter server for averaging and updating the model. Due to communication constraints, clients often use lossy compression techniques to compress the gradients, resulting in estimation inaccuracies. DME is more challenging when clients have diverse network conditions, such as constrained communication budgets and packet losses. In such settings, DME techniques often incur a significant increase in the estimation error leading to degraded learning performance. In this work, we propose a robust DME technique named EDEN that naturally handles heterogeneous communication budgets and packet losses. We derive appealing theoretical guarantees for EDEN and evaluate it empirically. Our results demonstrate that EDEN consistently improves over state-of-the-art DME techniques.

cs.LG↗

arXiv2022

Uniform Bounds for Scheduling with Job Size Estimates

We consider the problem of scheduling to minimize mean response time in M/G/1 queues where only estimated job sizes (processing times) are known to the scheduler, where a job of true size $s$ has estimated size in the interval $[βs, αs]$ for some $α\geq β> 0$. We evaluate each scheduling policy by its approximation ratio, which we define to be the ratio between its mean response time and that of Shortest Remaining Processing Time (SRPT), the optimal policy when true sizes are known. Our question: is there a scheduling policy that (a) has approximation ratio near 1 when $α$ and $β$ are near 1, (b) has approximation ratio bounded by some function of $α$ and $β$ even when they are far from 1, and (c) can be implemented without knowledge of $α$ and $β$? We first show that naively running SRPT using estimated sizes in place of true sizes is not such a policy: its approximation ratio can be arbitrarily large for any fixed $β< 1$. We then provide a simple variant of SRPT for estimated sizes that satisfies criteria (a), (b), and (c). In particular, we prove its approximation ratio approaches 1 uniformly as $α$ and $β$ approach 1. This is the first result showing this type of convergence for M/G/1 scheduling. We also study the Preemptive Shortest Job First (PSJF) policy, a cousin of SRPT. We show that, unlike SRPT, naively running PSJF using estimated sizes in place of true sizes satisfies criteria (b) and (c), as well as a weaker version of (a).

cs.PF↗

arXiv2022

The Supermarket Model with Known and Predicted Service Times

The supermarket model refers to a system with a large number of queues, where new customers choose d queues at random and join the one with the fewest customers. This model demonstrates the power of even small amounts of choice, as compared to simply joining a queue chosen uniformly at random, for load balancing systems. In this work we perform simulation-based studies to consider variations where service times for a customer are predicted, as might be done in modern settings using machine learning techniques or related mechanisms. Our primary takeaway is that using even seemingly weak predictions of service times can yield significant benefits over blind First In First Out queueing in this context. However, some care must be taken when using predicted service time information to both choose a queue and order elements for service within a queue; while in many cases using the information for both choosing and ordering is beneficial, in many of our simulation settings we find that simply using the number of jobs to choose a queue is better when using predicted service times to order jobs in a queue. In our simulations, we evaluate both synthetic and real-world workloads--in the latter, service times are predicted by machine learning. Our results provide practical guidance for the design of real-world systems; moreover, we leave many natural theoretical open questions for future work, validating their relevance to real-world situations.

cs.PF↗