東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Size-aware scheduling of TCP flows.

Guo, Liang.

FindBook

Google Book

Amazon

博客來

Size-aware scheduling of TCP flows.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Size-aware scheduling of TCP flows./
作者:	Guo, Liang.
面頁冊數:	135 p.
附註:	Source: Dissertation Abstracts International, Volume: 64-05, Section: B, page: 2264.
Contained By:	Dissertation Abstracts International64-05B.
標題:	Computer Science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3090408

Size-aware scheduling of TCP flows.
Guo, Liang.

Size-aware scheduling of TCP flows. - 135 p.

Source: Dissertation Abstracts International, Volume: 64-05, Section: B, page: 2264.

Thesis (Ph.D.)--Boston University, 2003.

Internet flow sizes are skewed: there are many small sized flows (or "mice") and a few extremely large flows (or "elephants"). However, most of the Internet traffic is due to these elephant flows. Previous job scheduling studies show that with such skewed distribution, a scheduling policy such as Shortest Job First (SJF), that gives high priority to short jobs, can significantly reduce the overall average response time with almost no penalty to large jobs.Subjects--Topical Terms:

626642
Computer Science.

Size-aware scheduling of TCP flows.
LDR:03292nmm 2200289 4500 001 1863439
005 20041216101837.5
008 130614s2003 eng d
035 $a (UnM)AAI3090408
035 $a AAI3090408
040 $a UnM $c UnM
100 1 $a Guo, Liang. $3 1926882
245 1 0 $a Size-aware scheduling of TCP flows.
300 $a 135 p.
500 $a Source: Dissertation Abstracts International, Volume: 64-05, Section: B, page: 2264.
500 $a Major Professor: Ibrahim Matta.
502 $a Thesis (Ph.D.)--Boston University, 2003.
520 $a Internet flow sizes are skewed: there are many small sized flows (or "mice") and a few extremely large flows (or "elephants"). However, most of the Internet traffic is due to these elephant flows. Previous job scheduling studies show that with such skewed distribution, a scheduling policy such as Shortest Job First (SJF), that gives high priority to short jobs, can significantly reduce the overall average response time with almost no penalty to large jobs.
520 $a The primary contribution of this thesis is the design, modeling, and performance evaluation of a scheduling policy that mimics SJF scheduling by implementing additional functionality at Internet routers. The focus is on scheduling flows controlled by the popular Transmission Control Protocol (TCP). Since routers do not have advanced knowledge of flow sizes, the scheme simply monitors incoming packets/bytes and assigns lower priority to flows that have already sent more than a certain amount of packets.
520 $a Two analytical models are developed in this thesis. One uses the Multi-Level (ML) feedback queueing model to characterize the threshold-based prioritization scheme, under the assumption of ideal bandwidth sharing among flows. An approximate analysis of this ML model is given to derive closed-form solutions to the average response time. The second model uses a numerical method to analyze a more realistic but less tractable bandwidth sharing model achieved by TCP. These models serve as stepping-stones for the development of an adaptive controller for parameterizing the proposed control system. The analytical and numerical results are validated with packet-level simulations and testbed experiments. The experimental prototype is implemented over the Linux netfilter Application Programming Interface over a wide-area network testbed.
520 $a The proposed size-aware differentiated scheduling scheme fits into a scalable architecture of Internet Traffic Managers (ITMs), where routers with enhanced functionalities (the ITMs) are deployed at strategic places in the Internet. Practical implementation issues such as incremental deployment and scalability of the size-aware scheduling through ITMs are also studied in this thesis. A random sampling mechanism is developed to reduce the implementation complexity of the threshold-based flow classification scheme at the ITMs. The effectiveness of the sampling mechanism and incremental deployment are highlighted by simulation results.
590 $a School code: 0017.
650 4 $a Computer Science. $3 626642
690 $a 0984
710 2 0 $a Boston University. $3 1017454
773 0 $t Dissertation Abstracts International $g 64-05B.
790 1 0 $a Matta, Ibrahim, $e advisor
790 $a 0017
791 $a Ph.D.
792 $a 2003
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3090408