Abstract
We consider the model of online computation with advice (Emek et al., Theor. Comput. Sci. 412(24): 2642–2656, 2011). In particular, we study the k-server problem under this model. We prove three upper bounds for this problem. First, we show a \(\lceil\frac{\lceil\log k\rceil}{b-2}\rceil\)-competitive online algorithm for general metric spaces with b bits of advice per request, where 3≤b≤logk. This improves upon the result of Böckenhauer et al. (ICALP (1), Lecture Notes in Computer Science, vol. 6755, pp. 207–218, 2011). Moreover, we believe that our algorithm and our analysis are more intuitive and simpler than those of Böckenhauer et al. Second, we give a 1-competitive online algorithm for finite trees which uses 2+2⌈log(p+1)⌉ bits of advice per request, where p is the caterpillar dimension of the tree. Lastly, we present a variant of the algorithm for the tree that is optimal for the line with 1-bit of advice.



Similar content being viewed by others
References
Böckenhauer, H.J., Komm, D., Královic, R., Královic, R., Mömke, T.: On the advice complexity of online problems. In: Dong, Y., Du, D.Z., Ibarra, O.H. (eds.) ISAAC. Lecture Notes in Computer Science, vol. 5878, pp. 331–340. Springer, Berlin (2009)
Böckenhauer, H.J., Komm, D., Královic, R., Královic, R.: On the advice complexity of the k-server problem. In: Aceto, L., Henzinger, M., Sgall, J. (eds.) ICALP (1). Lecture Notes in Computer Science, vol. 6755, pp. 207–218. Springer, Berlin (2011). Also as technical report at ftp://ftp.inf.ethz.ch/pub/publications/tech-reports/7xx/703.pdf
Borodin, A., El-Yaniv, R.: Online Computation and Competitive Analysis. Cambridge University Press, New York (1998)
Borodin, A., Linial, N., Saks, M.E.: An optimal online algorithm for metrical task systems. J. ACM 39, 373–382 (1992)
Dobrev, S., Královič, R., Pardubská, D.: How much information about the future is needed? In: SOFSEM’08: Proceedings of the 34th Conference on Current Trends in Theory and Practice of Computer Science, pp. 247–258. Springer, Berlin (2008)
Emek, Y., Fraigniaud, P., Korman, A., Rosén, A.: Online computation with advice. Theor. Comput. Sci. 412(24), 2642–2656 (2011)
Hromkovic, J., Královic, R., Královic, R.: Information complexity of online problems. In: Hlinený, P., Kucera, A. (eds.) MFCS. Lecture Notes in Computer Science, vol. 6281, pp. 24–36. Springer, Berlin (2010)
Komm, D., Královic, R.: Advice complexity and barely random algorithms. In: Cerná, I., Gyimóthy, T., Hromkovic, J., Jeffery, K.G., Královic, R., Vukolic, M., Wolf, S. (eds.) SOFSEM. Lecture Notes in Computer Science, vol. 6543, pp. 332–343. Springer, Berlin (2011)
Matoušek, J.: On embedding trees into uniformly convex Banach spaces. Isr. J. Math. 114, 221–237 (1999)
Acknowledgements
We thank Yuval Emek, Pierre Fraigniaud, Amos Korman, and Manor Mendel for useful discussions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Research supported in part by ANR projects QRAC and NeTOC.
Appendix: Minimum Matching
Appendix: Minimum Matching
Lemma 3
Let A and B be two sets of points of equal size on the line. For any a∈A there is a minimum weighted matching between the points of A and the points of B such that a is matched to a point b∈B, where b is the first element of B to the right or the left of a.
Proof
Given a minimum weighted matching between the sets A and B on the line, let a∈A be matched to a server e∈B such that e is not the first element of B to the right or the left of a and let b∈B be a point that is the first element of B to the left or the right of a and between a and e on the line. Let c∈A be the point to which b is matched. Let d(x,y) be the distance between the points x and y on the line. Without loss of generality, assume that b and e are to the right of a.
If c is between b and e, then d(a,b)+d(c,e)≤d(a,e)+d(b,c), so a can be matched to b and c can be matched to e without increasing the cost of the matching.
If c is to the right of e, then d(a,b)+d(e,c)≤d(a,e)+d(b,c) since d(a,b)+2d(b,e)+d(e,c)=d(a,e)+d(b,c). So, a can be matched to b and c can be matched to e without increasing the cost of the matching.
If c is to the left of a, then d(a,b)+d(c,e)=d(c,b)+d(a,e) since d(a,b)+d(c,e)=d(a,e)+2d(a,b)+d(c,b)=d(a,b)+d(c,e). So, a can be matched to b and c can be matched to e without increasing the cost of the matching.
The final case is that c is between a and b. Note that this is the previous case with a and c swapped. As in the previous case, a can be matched to b and c can be matched to e without increasing the cost of the matching. □
Rights and permissions
About this article
Cite this article
Renault, M.P., Rosén, A. On Online Algorithms with Advice for the k-Server Problem. Theory Comput Syst 56, 3–21 (2015). https://doi.org/10.1007/s00224-012-9434-z
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00224-012-9434-z