TY - GEN
T1 - Hardware-software co-design for network performance measurement
AU - Narayana, Srinivas
AU - Sivaraman, Anirudh
AU - Nathan, Vikram
AU - Alizadeh, Mohammad
AU - Walker, David P.
AU - Rexford, Jennifer L.
AU - Jeyakumar, Vimalkumar
AU - Kim, Changhoon
N1 - Funding Information:
We are grateful to the anonymous Hot-Nets reviewers, Hari Balakrishnan, Mina Tahmasbi, and Vibhaalakshmi Sivaraman for their thoughtful feedback on this paper. This work was funded in part by NSF grants CNS-1563826, CNS-1617702, CNS-1526791, and AiTF-1535948; DARPA grant HR0011-15-2-0047; and a gift from Cisco. We also thank the industrial members of the MIT Center for Wireless Networks and Mobile Computing (Wireless@ MIT) for their support.
PY - 2016/11/9
Y1 - 2016/11/9
N2 - Diagnosing performance problems in networks is important, for example to determine where packets experience high latency. However, existing diagnostic tools are constrained by limited switch mechanisms for measurement. As a result, operators use endpoint information to indirectly infer root causes for performance issues. Instead of designing piecemeal solutions to work around limited switch mechanisms, we believe that the right approach is to co-design language abstractions and switch hardware primitives for performance measurement. This approach provides confidence that the switch primitives are useful for a variety of existing and unanticipated use cases. We present a declarative query language that allows operators to ask a diverse set of network performance questions. We show that these queries can be implemented efficiently in switch hardware using a programmable key-value store primitive. Our preliminary evaluations show that our hardware design incurs modest additional chip area relative to existing switching chips, suggesting that it is a practical solution for network performance measurement.
AB - Diagnosing performance problems in networks is important, for example to determine where packets experience high latency. However, existing diagnostic tools are constrained by limited switch mechanisms for measurement. As a result, operators use endpoint information to indirectly infer root causes for performance issues. Instead of designing piecemeal solutions to work around limited switch mechanisms, we believe that the right approach is to co-design language abstractions and switch hardware primitives for performance measurement. This approach provides confidence that the switch primitives are useful for a variety of existing and unanticipated use cases. We present a declarative query language that allows operators to ask a diverse set of network performance questions. We show that these queries can be implemented efficiently in switch hardware using a programmable key-value store primitive. Our preliminary evaluations show that our hardware design incurs modest additional chip area relative to existing switching chips, suggesting that it is a practical solution for network performance measurement.
UR - http://www.scopus.com/inward/record.url?scp=85002194787&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85002194787&partnerID=8YFLogxK
U2 - 10.1145/3005745.3005775
DO - 10.1145/3005745.3005775
M3 - Conference contribution
AN - SCOPUS:85002194787
T3 - HotNets 2016 - Proceedings of the 15th ACM Workshop on Hot Topics in Networks
SP - 190
EP - 196
BT - HotNets 2016 - Proceedings of the 15th ACM Workshop on Hot Topics in Networks
PB - Association for Computing Machinery, Inc
T2 - 15th ACM Workshop on Hot Topics in Networks, HotNets 2016
Y2 - 9 November 2016 through 10 November 2016
ER -