Open AccessPosted Content
Efficient Distributed Workload (Re-)Embedding
TL;DR: A fundamental model which captures the tradeoff between the benefits and costs of dynamically collocating communication partners on servers, in an online manner is studied and a distributed online algorithm which is asymptotically almost optimal, i.e., almost matches the lower bound on the competitive ratio of any (distributed or centralized) online algorithm is derived.
read more
Abstract: Modern networked systems are increasingly reconfigurable, enabling demand-aware infrastructures whose resources can be adjusted according to the workload they currently serve. Such dynamic adjustments can be exploited to improve network utilization and hence performance, by moving frequently interacting communication partners closer, e.g., collocating them in the same server or datacenter. However, dynamically changing the embedding of workloads is algorithmically challenging: communication patterns are often not known ahead of time, but must be learned. During the learning process, overheads related to unnecessary moves (i.e., re-embeddings) should be minimized. This paper studies a fundamental model which captures the tradeoff between the benefits and costs of dynamically collocating communication partners on $\ell$ servers, in an online manner. Our main contribution is a distributed online algorithm which is asymptotically almost optimal, i.e., almost matches the lower bound (also derived in this paper) on the competitive ratio of any (distributed or centralized) online algorithm. As an application, we show that our algorithm can be used to solve a distributed union find problem in which the sets are stored across multiple servers.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures
Figure 1: An illustration of the model we consider. In the picture there are ` = 4 servers each depicted by a blue circle. Vertices assigned to a server are represented by black dots whereas white dots represent unused server capacities. Note that there are n = 24 vertices and each server has capacity (1 + ε)n/` = 8 for ε = 1/3. In the picture, server S0 has load 5 and server S1 has load 8. When two vertices communicated, we draw an edge between them depicted by a black line. Observe how this naturally gives rise to connected components and note that S1 contains a ground truth component of size n/` = 6. If the adversary were to insert the edge (u, v) next, the algorithm could, for example, move the connected component containing v to S3 at cost 2α. 
Figure 3: An illustration of the bipartition tree T for servers S0, . . . , S7. The internal nodes of the bipartition tree are denoted w1, . . . , w7. We highlighted the subtrees T = T (w1), T (w3), and T (w7). Here we obtain the server sets S(w1) = {S0, . . . , S7}, S(w3) = {S4, . . . , S7}, and S(w7) = {S6, S7}. 
Figure 2: The initial assignment considered in the proof of Lemma 3. The green and grey areas of the servers correspond to subsets of V0 and V1. Server S0 (S1) contains most of the vertices from V0 (V1). Here, OPT would move the green part from S1 to S0 and the grey part from S0 to S1.
Citations
Demand-Aware Network Design with Minimal Congestion and Route Lengths
Chen Avin,Kaushik Mondal,Stefan Schmid +2 more
- 01 Apr 2019
TL;DR: This paper presents the first bounded-degree, demand-aware network, ct-DAN, which minimizes both congestion and route lengths and is provably (asymptotically) optimal in each dimension individually.
51
Dynamic Balanced Graph Partitioning
TL;DR: It is proved that any deterministic online algorithm has a competitive ratio of at least $k$, even with significant augmentation, which is a constant competitive algorithm for the maximum matching variant.
20
Coupling VNF Orchestration and SDN Virtual Network Reconfiguration
Nemanja Deric,Amir Varasteh,Arsany Basta,Andreas Blenk,Rastin Pries,Michael Jarschel,Wolfgang Kellerer +6 more
- 18 Mar 2019
TL;DR: This demo presents the benefits of migrating a firewall VNF to a server, which is closer to its user, at runtime and shows how the migration is supported in a virtualized SDN environment.
Efficient Distributed Workload (Re-)Embedding
Monika Henzinger,Stefan Neumann,Stefan Schmid +2 more
- 26 Mar 2019
TL;DR: This paper studies a fundamental model which captures the tradeoff between the benefits and costs of dynamically collocating communication partners on l servers, in an online manner and produces a distributed online algorithm which is asymptotically almost optimal, i.e., almost matches the lower bound on the competitive ratio of any (distributed or centralized) online algorithm.
12
•Proceedings Article
Tight Bounds for Online Graph Partitioning
Monika Henzinger,Stefan Neumann,Harald Räcke,Stefan Schmid +3 more
- 10 Jan 2021
TL;DR: An improved lower bound as well as a deterministic polynomial-time online algorithm, that is asymptotically optimal, and an upper bound of $O(\log \ell + \log k)$ on its competitive ratio and show that no randomized online algorithm can achieve a competitive ratio of less than $Omega$.
12
References
•Book
Computers and Intractability: A Guide to the Theory of NP-Completeness
Michael Randolph Garey,David S. Johnson +1 more
- 01 Jan 1979
TL;DR: The second edition of a quarterly column as discussed by the authors provides a continuing update to the list of problems (NP-complete and harder) presented by M. R. Garey and myself in our book "Computers and Intractability: A Guide to the Theory of NP-Completeness,” W. H. Freeman & Co., San Francisco, 1979.
P4: programming protocol-independent packet processors
Pat Bosshart,Daniel P. Daly,Glen Gibb,Martin J. Izzard,Nick McKeown,Jennifer Rexford,Cole Schlesinger,Daniel Talayco,Amin Vahdat,George Varghese,David Walker +10 more
- 28 Jul 2014
TL;DR: This paper proposes P4 as a strawman proposal for how OpenFlow should evolve in the future, and describes how to use P4 to configure a switch to add a new hierarchical label.
Amortized efficiency of list update and paging rules
TL;DR: This article shows that move-to-front is within a constant factor of optimum among a wide class of list maintenance rules, and analyzes the amortized complexity of LRU, showing that its efficiency differs from that of the off-line paging rule by a factor that depends on the size of fast memory.
2.5K
Network traffic characteristics of data centers in the wild
Theophilus Benson,Aditya Akella,David A. Maltz +2 more
- 01 Nov 2010
TL;DR: An empirical study of the network traffic in 10 data centers belonging to three different categories, including university, enterprise campus, and cloud data centers, which includes not only data centers employed by large online service providers offering Internet-facing applications but also data centers used to host data-intensive (MapReduce style) applications.