A Simple yet Exact Analysis of the MultiQueue

Stefan Walzer,Marvin Williams

2024-10-11

Abstract:The MultiQueue is a relaxed concurrent priority queue consisting of $n$ internal priority queues, where an insertion uses a random queue and a deletion considers two random queues and deletes the minimum from the one with the smaller minimum. The rank error of the deletion is the number of smaller elements in the MultiQueue. Alistarh et al. [2] have demonstrated in a sophisticated potential argument that the expected rank error remains bounded by $O(n)$ over long sequences of deletions. In this paper we present a simpler analysis by identifying the stable distribution of an underlying Markov chain and with it the long-term distribution of the rank error exactly. Simple calculations then reveal the expected long-term rank error to be $\tfrac{5}{6}n-1+\tfrac{1}{6n}$. Our arguments generalize to deletion schemes where the probability to delete from a given queue depends only on the rank of the queue. Specifically, this includes deleting from the best of $c$ randomly selected queues for any $c>1$.

Data Structures and Algorithms

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper mainly discusses the performance analysis problem of **MultiQueue**, especially its **rank error** in long - time operation sequences. Specifically, the paper aims to provide a simpler and more accurate method to analyze the behavior of MultiQueue in a concurrent environment, especially the rank error during deletion operations.

#### Background and Motivation

1. **The need for concurrent priority queues**:
- The parallelism of modern computing hardware promotes the design of concurrent priority queues that allow multiple processing units to insert and delete elements simultaneously.
- Strict concurrent priority queues (such as linearizable priority queues) have the problem of poor scalability due to competition for the minimum element.

2. **Relaxing the concept of priority queues**:
- Relaxed priority queues allow deletion operations to not always extract the minimum element, but tend to extract smaller elements.
- This relaxation can improve scalability, although it may introduce additional workload (i.e., rank error).

3. **The design of MultiQueue**:
- MultiQueue is a relaxed priority queue using the "choose - one - of - two" strategy. It randomly assigns insertion operations to multiple internal priority queues and selects the smaller minimum value from two random queues for deletion.
- This design improves scalability and robustness.

#### Main contributions of the paper

1. **Simplified analysis method**:
- The author proposes an analysis method that is simpler and more accurate than the potential argument of Alistarh et al.
- By identifying the stable distribution of the underlying Markov chain, the author can accurately describe the distribution of long - term rank error.

2. **Accurate long - term rank error formula**:
- For the case of \( c = 2 \), the author derives the expected long - term rank error as:
\[
E[E]=\frac{5}{6}n - 1+\frac{1}{6n}
\]
- This formula is not only concise but also accurate and is applicable to any \( n\in\mathbb{N} \).

3. **Generalized deletion scheme**:
- The analysis method can be generalized to all deletion schemes that depend on queue rank probabilities, including deletion from any number \( c > 1 \) of randomly selected queues.

4. **Independent analysis of related processes**:
- The author also analyzes a related exponential - jump process inspired by Alistarh et al., further verifying the effectiveness of the method.

#### Conclusion

The paper provides a new and simpler analysis method by introducing the Markov chain model and geometric distribution, and successfully solves the long - term rank error problem of MultiQueue. This method not only simplifies the previous complex analysis but also provides more accurate results, which helps to better understand the behavior and performance advantages of relaxed priority queues.