Please use this identifier to cite or link to this item:
Title: Parallel analytics as a service
Authors: Wong, P
He, Z
Lo, E 
Issue Date: 2013
Source: SIGMOD '13 Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data, New York, Jun 22 - 27, 2013, p. 25-36
Abstract: By providing the best-effort service, the Internet Protocol (IP) does not maintain the same order of packets sent out by a host. Therefore, due to the route change, parallelism inside a switch, and load-balancing schemes, IP packets can be received in an order different from the original one. Such packet reordering events could cause serious performance degradation in TCP and UDP applications. As a result, a number of measurement methods have recently been proposed to enable any Internet host to detect packet reordering from itself to another host. However, these methods have encountered a number of practical difficulties, such as rate-limiting and filtering imposed on ICMP and TCP SYN packets. Moreover, some of the methods cannot detect packet reordering in all scenarios. In this paper we present three new methods for end-to-end packet reordering measurement. Since these methods are based on the TCP data channel, the probing and response messages will not be affected by any intermediaries on an Internet path. We have validated and tested the methods in 20 most common systems and implemented them in a tool called POINTER. We also present measurement results obtained from 200 websites in the Internet.
Keywords: Cloud databases
Multi-tenant databases
Parallel databases
ISBN: 978-1-4503-2037-5
DOI: 10.1145/2463676.2463714
Appears in Collections:Conference Paper

View full-text via PolyU eLinks SFX Query
Show full item record


Last Week
Last month
Citations as of Aug 18, 2020

Page view(s)

Last Week
Last month
Citations as of Sep 29, 2020

Google ScholarTM



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.