Diluting the Scalability Boundaries: Exploring the Use of Disaggregated Architectures for High-Level Network Data Analysis

Traditional data centers are designed with a rigid architecture of fit-for-purpose servers that provision resources beyond the average workload in order to deal with occasional peaks of data. Heterogeneous data centers are pushing towards more cost-efficient architectures with better resource provisioning. In this paper we study the feasibility of using disaggregated architectures for intensive data applications, in contrast to the monolithic approach of server-oriented architectures. Particularly, we have tested a proactive network analysis system in which the workload demands are highly variable. In the context of the dReDBox disaggregated architecture, the results show that the overhead caused by using remote memory resources is significant, between 66% and 80%, but we have also observed that the memory usage is one order of magnitude higher for the stress case with respect to average workloads. Therefore, dimensioning memory for the worst case in conventional systems will result in a notable waste of resources. Finally, we found that, for the selected use case, parallelism is limited by memory. Therefore, using a disaggregated architecture will allow for increased parallelism, which, at the same time, will mitigate the overhead caused by remote memory.

[1]  Mohsen Guizani,et al.  Internet of Things: A Survey on Enabling Technologies, Protocols, and Applications , 2015, IEEE Communications Surveys & Tutorials.

[2]  J. Manyika,et al.  Disruptive technologies: Advances that will transform life, business, and the global economy , 2013 .

[3]  Paul M. Carpenter,et al.  EUROSERVER: Energy Efficient Node for European Micro-Servers , 2014, 2014 17th Euromicro Conference on Digital System Design.

[4]  K. Hasharoni,et al.  High BW Parallel Optical Interconnects , 2014 .

[5]  Salvatore Spadaro,et al.  On the benefits of resource disaggregation for virtual data centre provisioning in optical data centres , 2017, Comput. Commun..

[6]  Ioannis Tomkos,et al.  A Survey on Optical Interconnects for Data Centers , 2012, IEEE Communications Surveys & Tutorials.

[7]  Tzi-cker Chiueh,et al.  Marlin: A memory-based rack area network , 2014, 2014 ACM/IEEE Symposium on Architectures for Networking and Communications Systems (ANCS).

[8]  Richard E. Brown,et al.  United States Data Center Energy Usage Report , 2016 .

[9]  Thomas F. Wenisch,et al.  Disaggregated memory for expansion and sharing in blade servers , 2009, ISCA '09.

[10]  Lena Wosinska,et al.  Energy-Efficient Elastic Optical Interconnect Architecture for Data Centers , 2014, IEEE Communications Letters.

[11]  Victor I. Chang,et al.  Composable architecture for rack scale big data computing , 2017, Future Gener. Comput. Syst..

[12]  Kostas Katrinis,et al.  On interconnecting and orchestrating components in disaggregated data centers: The dReDBox project vision , 2016, 2016 European Conference on Networks and Communications (EuCNC).

[13]  Javier Aracil,et al.  Multi‐granular, multi‐purpose and multi‐Gb/s monitoring on off‐the‐shelf systems , 2014, Int. J. Netw. Manag..

[14]  Mateo Valero,et al.  Supercomputing with commodity CPUs: Are mobile SoCs ready for HPC? , 2013, 2013 SC - International Conference for High Performance Computing, Networking, Storage and Analysis (SC).

[15]  Maurizio Martinelli,et al.  nDPI: Open-source high-speed deep packet inspection , 2014, 2014 International Wireless Communications and Mobile Computing Conference (IWCMC).

[16]  Mario Nemirovsky,et al.  Disaggregated Computing. An Evaluation of Current Trends for Datacentres , 2017, ICCS.

[17]  Javier Aracil,et al.  Multi-Gbps HTTP Traffic Analysis in Commodity Hardware Based on Local Knowledge of TCP Streams , 2017, Comput. Networks.

[18]  Chung-Sheng Li,et al.  Disaggregated and optically interconnected memory: when will it be cost effective? , 2015, ArXiv.

[19]  Albert G. Greenberg,et al.  Towards a next generation data center architecture: scalability and commoditization , 2008, PRESTO '08.

[20]  Gustavo Sutter,et al.  Automated synthesis of FPGA-based packet filters for 100 Gbps network monitoring applications , 2016, 2016 International Conference on ReConFigurable Computing and FPGAs (ReConFig).

[21]  Michael Mesh,et al.  A 1.3 Tb/s parallel optics VCSEL link , 2014, Photonics West - Optoelectronic Materials and Devices.

[22]  Pawel Gburzynski,et al.  Load balancing for parallel forwarding , 2005, IEEE/ACM Transactions on Networking.

[23]  Scott Shenker,et al.  Network support for resource disaggregation in next-generation datacenters , 2013, HotNets.