NFSlicer: Data Movement Optimization for Shallow Network Functions

NFSlicer: Data Movement Optimization for Shallow Network Functions. Sarma, A., Seyedroudbari, H., Gupta, H., Ramachandran, U., & Daglis, A.
abstract bibtex

Network Function (NF) deployments on commodity servers have become ubiquitous in datacenters and enterprise settings. Many commonly used NFs such as ﬁrewalls, load balancers and NATs are shallow—i.e., they only examine the packet’s header, despite the entire packet being transferred on and off the server. As a result, the gap between moved and inspected data when handling large packets exceeds 20×. At modern network rates, such excess data movement is detrimental to performance, hurting both the average and 90% tail latency of large packets by up to 1.7×. Our thorough performance analysis identiﬁes high contention on the NIC-server PCIe interface and in the server’s memory hierarchy as the main bottlenecks.

@article{sarma_nfslicer_nodate,
	title = {{NFSlicer}: {Data} {Movement} {Optimization} for {Shallow} {Network} {Functions}},
	abstract = {Network Function (NF) deployments on commodity servers have become ubiquitous in datacenters and enterprise settings. Many commonly used NFs such as ﬁrewalls, load balancers and NATs are shallow—i.e., they only examine the packet’s header, despite the entire packet being transferred on and off the server. As a result, the gap between moved and inspected data when handling large packets exceeds 20×. At modern network rates, such excess data movement is detrimental to performance, hurting both the average and 90\% tail latency of large packets by up to 1.7×. Our thorough performance analysis identiﬁes high contention on the NIC-server PCIe interface and in the server’s memory hierarchy as the main bottlenecks.},
	language = {en},
	author = {Sarma, Anirudh and Seyedroudbari, Hamed and Gupta, Harshit and Ramachandran, Umakishore and Daglis, Alexandros},
}

Downloads: 0

{"_id":"mn8hFhfXktqJnrTem","bibbaseid":"sarma-seyedroudbari-gupta-ramachandran-daglis-nfslicerdatamovementoptimizationforshallownetworkfunctions","author_short":["Sarma, A.","Seyedroudbari, H.","Gupta, H.","Ramachandran, U.","Daglis, A."],"bibdata":{"bibtype":"article","type":"article","title":"NFSlicer: Data Movement Optimization for Shallow Network Functions","abstract":"Network Function (NF) deployments on commodity servers have become ubiquitous in datacenters and enterprise settings. Many commonly used NFs such as ﬁrewalls, load balancers and NATs are shallow—i.e., they only examine the packet’s header, despite the entire packet being transferred on and off the server. As a result, the gap between moved and inspected data when handling large packets exceeds 20×. At modern network rates, such excess data movement is detrimental to performance, hurting both the average and 90% tail latency of large packets by up to 1.7×. Our thorough performance analysis identiﬁes high contention on the NIC-server PCIe interface and in the server’s memory hierarchy as the main bottlenecks.","language":"en","author":[{"propositions":[],"lastnames":["Sarma"],"firstnames":["Anirudh"],"suffixes":[]},{"propositions":[],"lastnames":["Seyedroudbari"],"firstnames":["Hamed"],"suffixes":[]},{"propositions":[],"lastnames":["Gupta"],"firstnames":["Harshit"],"suffixes":[]},{"propositions":[],"lastnames":["Ramachandran"],"firstnames":["Umakishore"],"suffixes":[]},{"propositions":[],"lastnames":["Daglis"],"firstnames":["Alexandros"],"suffixes":[]}],"bibtex":"@article{sarma_nfslicer_nodate,\n\ttitle = {{NFSlicer}: {Data} {Movement} {Optimization} for {Shallow} {Network} {Functions}},\n\tabstract = {Network Function (NF) deployments on commodity servers have become ubiquitous in datacenters and enterprise settings. Many commonly used NFs such as ﬁrewalls, load balancers and NATs are shallow—i.e., they only examine the packet’s header, despite the entire packet being transferred on and off the server. As a result, the gap between moved and inspected data when handling large packets exceeds 20×. At modern network rates, such excess data movement is detrimental to performance, hurting both the average and 90\\% tail latency of large packets by up to 1.7×. Our thorough performance analysis identiﬁes high contention on the NIC-server PCIe interface and in the server’s memory hierarchy as the main bottlenecks.},\n\tlanguage = {en},\n\tauthor = {Sarma, Anirudh and Seyedroudbari, Hamed and Gupta, Harshit and Ramachandran, Umakishore and Daglis, Alexandros},\n}\n\n","author_short":["Sarma, A.","Seyedroudbari, H.","Gupta, H.","Ramachandran, U.","Daglis, A."],"key":"sarma_nfslicer_nodate","id":"sarma_nfslicer_nodate","bibbaseid":"sarma-seyedroudbari-gupta-ramachandran-daglis-nfslicerdatamovementoptimizationforshallownetworkfunctions","role":"author","urls":{},"metadata":{"authorlinks":{}},"html":""},"bibtype":"article","biburl":"https://api.zotero.org/users/8338458/collections/FNZYTNHX/items?key=UHPrke7WIAIZpti8nDK8xzkr&format=bibtex&limit=100","dataSources":["TMqSANL7xJruPR7m7"],"keywords":[],"search_terms":["nfslicer","data","movement","optimization","shallow","network","functions","sarma","seyedroudbari","gupta","ramachandran","daglis"],"title":"NFSlicer: Data Movement Optimization for Shallow Network Functions","year":null}