Abstract
In the online balanced graph repartitioning problem, one has to maintain a clustering of n nodes into \(\ell \) clusters, each having \(k = n / \ell \) nodes. During runtime, an online algorithm is given a stream of communication requests between pairs of nodes: an inter-cluster communication costs one unit, while the intra-cluster communication is free. An algorithm can change the clustering, paying unit cost for each moved node.
This natural problem admits a simple \(O(\ell ^2 \cdot k^2)\)-competitive algorithm Comp, whose performance is far apart from the best known lower bound of \(\varOmega (\ell \cdot k)\). One of open questions is whether the dependency on \(\ell \) can be made linear; this question is of practical importance as in the typical datacenter application where virtual machines are clustered on physical servers, \(\ell \) is of several orders of magnitude larger than k. We answer this question affirmatively, proving that a simple modification of Comp is \((\ell \cdot 2^{O(k)})\)-competitive.
On the technical level, we achieve our bound by translating the problem to a system of linear integer equations and using Graver bases to show the existence of a “small” solution.
Supported by Polish National Science Centre grant 2016/22/E/ST6/00499, by Center for Foundations of Modern Computer Science (Charles University project UNCE/SCI/004), by the project 19-27871X of GA ČR, and by NSF CAREER grant 1651861.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In the reality, the network load generated by a single request is much smaller than the load of migrating a whole virtual machine. This has been captured by some papers, which assigned cost \(\alpha \ge 1\) to the latter event (\(\alpha \) is then a parameter of the problem). However, this additional difficulty can be resolved by standard rent-or-buy approaches (reacting only to every \(\alpha \)-th request between a given pair of nodes). Therefore, and also to keep the description simple, in this paper, we assume that \(\alpha = 1\).
- 2.
Deciding whether such remapping exists is NP-hard. As typical for online algorithms, however, our focus is on studying the disadvantage of not knowing the future rather than on computational complexity.
- 3.
One could also bound the size of a solution along the lines of Schrijver [14, Corollary 17.1b], which boils down to a determinant bound, same as our proof. In general, Graver basis elements may be much smaller, but in out specific case, the resulting bound would be asymptotically the same.
References
Andreev, K., Räcke, H.: Balanced graph partitioning. Theory Comput. Syst. 39(6), 929–939 (2006). https://doi.org/10.1007/s00224-006-1350-7
Avin, C., Bienkowski, M., Loukas, A., Pacut, M., Schmid, S.: Dynamic balanced graph partitioning. SIAM J. Disc. Math. 34(3), 1791–1812 (2020). https://doi.org/10.1137/17M1158513
Avin, C., Loukas, A., Pacut, M., Schmid, S.: Online balanced repartitioning. In: Proceedings 30th International Symposium on Distributed Computing (DISC), pp. 243–256 (2016). https://doi.org/10.1007/978-3-662-53426-7_18
Borodin, A., El-Yaniv, R.: Online Computation and Competitive Analysis. Cambridge University Press, Cambridge (1998)
Chowdhury, M., Zaharia, M., Ma, J., Jordan, M.I., Stoica, I.: Managing data transfers in computer clusters with Orchestra. In: ACM SIGCOMM, pp. 98–109 (2011). https://doi.org/10.1145/2018436.2018448
Erdős, P.: On an elementary proof of some asymptotic formulas in the theory of partitions. Ann. Math. 43(3), 437–450 (1942). https://doi.org/10.2307/1968802
Henzinger, M., Neumann, S., Räcke, H., Schmid, S.: Tight bounds for online graph partitioning. In: Proceedings of 32nd ACM-SIAM Symposium on Discrete Algorithms (SODA), pp. 2799–2818 (2021). https://doi.org/10.1137/1.9781611976465.166
Henzinger, M., Neumann, S., Schmid, S.: Efficient distributed workload (re-)embedding. In: 2019 SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems, pp. 43–44 (2019). https://doi.org/10.1145/3309697.3331503
Krauthgamer, R., Feige, U.: A polylogarithmic approximation of the minimum bisection. SIAM Rev. 48(1), 99–130 (2006). https://doi.org/10.1137/050640904
Onn, S.: Nonlinear discrete optimization. Zurich Lectures in Advanced Mathematics, European Mathematical Society (2010)
Pacut, M., Parham, M., Schmid, S.: Brief announcement: deterministic lower bound for dynamic balanced graph partitioning. In: Proceedings 39th ACM Symposium on Principles of Distributed Computing (PODC), pp. 461–463 (2020). https://doi.org/10.1145/3382734.3405696
Pacut, M., Parham, M., Schmid, S.: Optimal online balanced graph partitioning. In: Proceedings 40th IEEE International Conference on Computer Communications (INFOCOM), pp. 1–9 (2021). https://doi.org/10.1109/INFOCOM42981.2021.9488824
Räcke, H.: Optimal hierarchical decompositions for congestion minimization in networks. In: Proceedings of 40th ACM Symposium on Theory of Computing (STOC), pp. 255–264 (2008). https://doi.org/10.1145/1374376.1374415
Schrijver, A.: Theory of Linear and Integer Programming. John Wiley & Sons, Hoboken (1998)
Sleator, D.D., Tarjan, R.E.: Amortized efficiency of list update and paging rules. Commun. ACM 28(2), 202–208 (1985). https://doi.org/10.1145/2786.2793
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
Bienkowski, M., Böhm, M., Koutecký, M., Rothvoß, T., Sgall, J., Veselý, P. (2021). Improved Analysis of Online Balanced Clustering. In: Koenemann, J., Peis, B. (eds) Approximation and Online Algorithms. WAOA 2021. Lecture Notes in Computer Science(), vol 12982. Springer, Cham. https://doi.org/10.1007/978-3-030-92702-8_14
Download citation
DOI: https://doi.org/10.1007/978-3-030-92702-8_14
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-92701-1
Online ISBN: 978-3-030-92702-8
eBook Packages: Computer ScienceComputer Science (R0)