Abstract
Elasticity is very important to the scale-out distributed systems running on today’s large-scale multi-tenant clouds, regardless public or private. An elastic distributed data processing system must have the capability of: (1) dynamically balancing the computing load among workers due to their performance heterogeneity and dynamicity; (2) fast recovering the lost memory state of failure workers with acceptable overheads during the regular execution.
Unfortunately, we found that the design of the state-of-the-art distributed graph computing system only works well in small sized dedicated clusters. We implement a distributed graph computing prototype, X-Graph, and demonstrate the capabilities of being elastic in three ways. First, we present menger, a novel two-level graph partition framework, which further splits one worker-level partition into several sub-partitions as the basic migration units, and each has the “migration affinity” with one of the other workers. Second, we implement a dynamical load balancer based on menger, which prefers the worker that has the affinity of the sub-partition to be migrated as the destination, and completely avoids the costly sophistical graph re-partitioning algorithms. Third, we implement a differentiated replication frame-work, which supports parallel recovery for lost partitions just like general-purpose dataflow systems.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Giraph. https://giraph.apache.org/
Graphlab. http://graphlab.org/
Hadoop. http://hadoop.apache.org/
Ahmad, F., Chakradhar, S., Raghunathan, A., Vijaykumar, T.N.: Tarazu: optimizing mapreduce on heterogeneous clusters. In: ASPLOS 2012, pp. 61–74 (2012)
Chen, R., Weng, X., He, B., Yang, M., Choi, B., Li, X.: Improving large graph processing on partitioned graphs in the cloud. In: SoCC 2012, p. 3 (2012)
Cipar, J., Ho, Q., Kim, J.K., Lee, S., Ganger, G.R., Gibson, G.: Solving the straggler problem with bounded staleness. In: HotOS 2013 (2013)
Dean, J., Ghemawat, S.: MapReduce: simplified data processing on large clusters. Commun. ACM 51, 107–113 (2008)
Gonzalez, J.E., Low, Y., Gu, H., Bickson, D., Guestrin, C.: PowerGraph: distributed graph-parallel computation on natural graphs. In: OSDI 2012, p. 2 (2012)
Hendrickson, B., Devine, K.: Dynamic load balancing in computational mechanics. Comput. Methods Appl. Mech. Eng. 184, 485–500 (2000)
Hindman, B., Konwinski, A., Zaharia, M., Ghodsi, A., Joseph, A.D., Katz, R., Shenker, S., Stoica, I.: Mesos: a platform for fine-grained resource sharing in the data center. In: NSDI 2011, p. 22 (2011)
Khayyat, Z., Awara, K., Alonazi, A.: Mizan: a system for dynamic load balancing in large-scale graph processing. In: EuroSys 2013, pp. 169–182 (2013)
Kumar, V., Vavilapallih, Murihyh, A.C., Douglasm, C., Agarwali, S., Konarh, M., Evansy, R., Gravesy, T., Lowey, J., Shahh, H., Sethh, S., Sahah, B., Curinom, C., Omaleyh, O., Radiah, S.: Apache hadoop YARN: yet another resource negotiator. In: SoCC 2013, p. 5 (2013)
Kyrola, A., Blelloch, G., Guestrin, C.: GraphChi: large-scale graph computation on just a pc. In: OSDI 2012, pp. 31–46 (2012)
Low, Y., Gonzalez, J., Kyrola, A., Bickon, D., Guestrin, C., Hellersten, J.M.: GraphLab: a new framework for parallel machine learning. In: UAI 2010 (2010)
Low, Y., Gonzalez, J., Kyrola, A., Bickon, D., Guestrin, C., Hellersten, J.M.: Distributed GraphLab: a framework for machine learning in the cloud. In: PVLDB 2012, pp. 716–727 (2012)
Malewicz, G., Austern, M.H., L., Hundt, R.: Whare-Map: heterogeneity in homogeneous warehouse-scale computers. In: ISCA 2013, pp. 619–630 (2013)
Nguyen, D., Lenharth, A., Pingali, K.: A lightweight infrastructure for graph analytics. In: SOSP 2013, pp. 456–471 (2013)
Power, R., Li, J.: Piccolo: building fast, distributed programs with partitioned tables. In: OSDI 2010, pp. 1–14 (2010)
Roy, A., Mihailovic, I., Zwaenpoel, W.: X-Stream: edge-centric graph processing using streaming partitions. In: SOSP 2013, pp. 472–488 (2013)
Salihoglu, S., Widom, J.: GPS: a graph processing system. In: SSDBM 2013, p. 8 (2013)
Schwarzkopf, M., Konwinski, A., Abdelmalek, M., Wilkes, J.: Omega: flexible, scalable schedulers for large compute clusters. In: EuroSys 2013, pp. 351–364 (2013)
Stanton, I., Kliot, G.: Streaming graph partitioning for large distributed graphs. In: KDD 2012, pp. 1222–1230 (2012)
Yu, Y., Isard, M., Fetterly, D., Budiu, M., Lfarer-Lingsson, Kumar, P., Currey, G.J.: DryadLINQ: a system for general-purpose distributed data-parallel computing using a high-level language. In: OSDI 2008, pp. 1–14 (2008)
Zaharia, M., Chowdhury, M., Das, T., Dave, A., Ma, J., Mccauley, M., Frankin, M.J., Shenker, S., Stoca, I.: Resilient distributed datasets: a fault-tolerant abstraction for in-memory cluster computing. In: NSDI 2012 (2012)
Zhang, X., Tune, E., Hagmann, R., Jnagal, R., Gokhale, V., Wilkes, J.: CPI2: CPU performance isolation for shared compute cluster. In: EuroSys 2013, pp. 379–391 (2013)
Acknowledgement
This paper is partly supported by the NSFC under grant No. 61433019 and No. 61370104, International Science and Technology Cooperation Program of China under grant No. 2015DFE12860, MOE- Intel Special Research Fund of Information Technology under grant MOE-INTEL-2012-01, and Chinese Universities Scientific Fund under grant No. 2014TS008.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2015 Springer International Publishing Switzerland
About this paper
Cite this paper
Lu, L., Shi, X., Jin, H. (2015). Towards Truly Elastic Distributed Graph Computing in the Cloud. In: Yao, L., Xie, X., Zhang, Q., Yang, L., Zomaya, A., Jin, H. (eds) Advances in Services Computing. APSCC 2015. Lecture Notes in Computer Science(), vol 9464. Springer, Cham. https://doi.org/10.1007/978-3-319-26979-5_23
Download citation
DOI: https://doi.org/10.1007/978-3-319-26979-5_23
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-26978-8
Online ISBN: 978-3-319-26979-5
eBook Packages: Computer ScienceComputer Science (R0)