Center for Machine Learning and Intelligent Systems
About  Citation Policy  Donate a Data Set  Contact


Repository Web            Google
View ALL Data Sets

Optical Interconnection Network Data Set
Download: Data Folder, Data Set Description

Abstract: This dataset contains 640 performance measurements from a simulation of 2-Dimensional Multiprocessor Optical Interconnection Network.

Data Set Characteristics:  

Multivariate

Number of Instances:

640

Area:

Computer

Attribute Characteristics:

Integer, Real

Number of Attributes:

10

Date Donated

2018-03-29

Associated Tasks:

Classification, Regression

Missing Values?

N/A

Number of Web Hits:

8005


Source:

Cigdem Inan ACI- Mersin University, Dept of Computer Engineering, caci '@' mersin.edu.tr
Mehmet Fatih AKAY- Cukurova University, Dept of Computer Engineering, mfakay '@' cu.edu.tr


Data Set Information:

All simulations have done under the software named OPNET Modeler. Message passing is used as the communication mechanism in which any processor can submit to the network a point-to-point message destined at any other processor. M/M/1 queue is considered in the calculations which consist of a First-in First-Out buffer with packet arriving randomly according to a Poisson arrival process, and a processor, that retrieves packets from the buffer at a specified service rate. In all simulations, it is assumed that the processor at each node extracts a packet from an input queue, processes it for a period of time and when that period expires, it generates an output data message. The size of each input queue is assumed as infinite. A processor becomes idle only when all its input queues are empty.


Attribute Information:

The summary of the attributes is given below. Please read the paper ([Web Link]) for detailed descriptions of the attributes.

Node Number: The number of the nodes in the network. (8x8 or 4x4).

Thread Number: The number of threads in each node at the beginning of the simulation.

Spatial Distribution: The performance of the network is evaluated using synthetic traffic workloads. Uniform (UN), Hot Region (HR), Bit reverse (BR) and Perfect Shuffle (PS) traffic models have been included.

Temporal Distribution: Temporal distribution of packet generation is implemented by independent traffic sources. In our simulations, we utilized client–server traffic (i.e., a server node sends packets to respond to the reception of packets from clients) and asynchronous traffic (i.e., initially, all nodes generate traffic independently of the others; as time progresses, traffic generation at the source/destination nodes depends
on the receipt of messages from destination/source nodes).

T/R: Message transfer time (T ) Uniformly distributed with mean in range from 20 to 100 clock cycles. Thread run time (R) Exponentially distributed with a mean of 100 clock cycles.

Processor Utilization: The average processor utilization measures the percent of time that threads are running in the processor.

Channel Waiting Time: Average waiting time of a packet at the output channel queue until it is serviced by the channel.

Input Waiting Time: Average waiting time of a packet until it is serviced by the processor.

Network Response Time: The time between a request message is enqueued at the output channel and the corresponding data message is received in the input queue.

Channel Utilization: The percent of time that the channel is busy transferring packets to the network.







Relevant Papers:

M.F. Akay, C¸.I. Aci, F. Abut, Predicting the Performance Measures of a 2-Dimensional Message Passing Multiprocessor Architecture by Using Machine Learning Methods. Neural Network World 71(5):1907-1931. DOI: 10.14311/NNW.2015.25.013.



Citation Request:

Acı, Ç.İ. & Akay, M.F. A hybrid congestion control algorithm for broadcast-based architectures with multiple input queues. J Supercomput (2015) 71: 1907. [Web Link]


Supported By:

 In Collaboration With:

About  ||  Citation Policy  ||  Donation Policy  ||  Contact  ||  CML