CSC8530 - Sebastian Niezgoda

Title:

9/27 - Summary of P2P Networks by Sebastian Niezgoda (hosted by: Justin Wetherell)

Target Audience:

A novice and/or intermediate computer user.

Type of Document:

A general overview and summary of P2P with technical topics introduced as well.

Summary:

The article provides a very nice overview and introduction into peer to peer systems.  The site uses the more popular applications such as Napster and the GNUtella network to describe how P2P works.  Same examples are then extended to address the news covered music downloads and copyright infringement topics. The article then discusses how P2P Search Engines and P2P Routing works.  Finally, issues of security as relating to P2P are also addressed.

Conclusion:

The first thing that hit me about the article was the quote from CNN on the introductory page stating that peer to peer networks such as GNUtella are not going away because of their distribution - there is no central server to shut down or terminate.  Keeping this in mind while going through the history and current news ("Music and P2P" and "Copyright and P2P") I felt that the message sent to a novice computer user was that P2P needs not be fought to terminate but should be regulated for maximum benefit.

I would recommend the article to anybody looking for a general overview and introduction to P2P and for somebody who is looking to learn more about specific P2P applications or more technical issues such as routing or security.  I especially like the fact that each section of the article is an entity of its own with an introduction and sub-sections, which helps the reader navigate directly to learn about topics of their choice.

I personally enjoyed the section on P2P routing the most.  Learning about the packet headers and payloads, different protocols, etc. helped me visualize what actual P2P traffic looks like.  I found the visual aids very helpful as well.

Top

Title:

10/4 - Summary of Clusters, interconnect architecture, and applications by Sebastian Niezgoda (hosted by:  David Acker)

Target Audience:

Programmers and software developers.

Type of Document:

Article.

Summary:

The article discusses clusters. The author presents scalability of these applications as the largest drawback to using clusters. The underlying communication delays are the largest bottleneck of cluster systems. There are four areas based on which the degree to scalability could be determine. These include the amount and frequency of intra-node communication, type of communication (synchronous vs. asynchronous), and which nodes are communicating.

High Performance Computing (HPC) uses the Message Passing Interface (MPI) to execute communications. The article presents four most popular communication technologies discussing their advantages and disadvantages. These are Gigabit Ethernet, Myrinet by Myricom, QsNet by Quadrics, and SCI from Dolphin. The conclusion drawn is that the proprietary technologies, the three latter ones, are much faster than Gigabit Ethernet due to the large overhead required by that technology. The speed differences are not due to the Network Interface Cards, which usually perform the same way, but are affected by the nodes that are passing messages in the system.

Conclusion:

What I learned is why Gigabit Ethernet is slower than proprietary protocols. It's not so much the fact that it is slower that surprised me but why. I learned that each TCP/IP packet must have three copies existing in the main memory on a Linux system. This drawback requires that a single machine is dedicated on the network to solely moving packets along.

I would definitely recommend this article to those who may possibly write software for a cluster. It is not very technical or overly detailed but it provides enough information for a programmer to determine whether scalability on a network is something that needs to be addressed.

Top

Title:

10/18 - Summary of CORBA Overview by Sebastian Niezgoda (hosted by: David Vaglia)

Target Audience:

Programmers.

Type of Document:

An introductory article.

Summary:

CORBA, Common Object Request Broker Architecture, is introduced as a new distributed infrastructure.  It eases network programming by automating some tasks, error handling, marshalling and unmarshalling, etc.  OMG, the Object Management Group, is tasked with standardizing CORBA.  The Object Request Broker sits between Object Services and Application Interfaces/Domain Interfaces/Common Facilities serving as a messenger.

CORBA ORB is a communication layer of the framework, which serves as an abstraction interface for application interface communication.  When invoked by a client, ORB finds an available object and transparently allowing interaction between the client and the object.

Conclusion:

The overview quickly taught me about CORBA.  Most importantly, I now have an idea of how CORBA actually works.  The diagrams and short descriptions proved very useful and give a very good summary of parties involved (clients and servants, etc.) and how they interact.

I found this article useful and would recommend it to somebody looking to get a quick intro into CORBA.

Top

Title:

11/1 - Summary of Web Services Architectures: How They Stack Up by Sebastian Niezgoda (hosted by: David Vaglia)

Target Audience:

Web services developers.

Type of Document:

A comparison of three variations of web services architectures.

Summary:

The article compared three web service architecture frameworks from WebServices.org, IBM, and the W3C. Web services came about to provide interoperability. However, with each organization implementing different architecture stacks using different technologies, an agreement should be reached so that only most universal, most suitable one becomes the de-facto standard.

The article presents and summarizes the web services architecture stacks for the three implementations listed above. Each layer is identified and implementation tools (i.e. UDDI, SOAP, CORBA) are also listed.

The conclusion of the article suggests that "IBM's architecture is most acceptable." I believe this statement has been made because IBM's architecture stack is very clearly separated into logical layers. That is, negotiation is distinct from flow, description/publication, etc. whereas W3C's stack is very complicated (consisting of 3 parts) and WebServices.org does not appear to be as cleanly defined.

Conclusion:

This overview was not the easiest to read. Fortunately, the diagrams provide a quick and dirty introduction into the different architectures being discussed. The article was definitely meant for developers who are looking into implementing web services who are yet to make the decision as to which way to proceed. This is the message I took away: choose wisely because although web services are meant to work together, the actual implementations differ and a smart decision up front (as to which path should be followed) will most likely save trouble down the road as these architectures are still developing.

Top

Title:

11/1 - Summary of Distributed database by Sebastian Niezgoda (hosted by: Ducson Ngyuen)

Target Audience:

Novice computer users.

Type of Document:

An encyclopedia article, a very brief introduction into distributed databases.

Summary:

The article is an entry from the Wikipedia encyclopedia. Thus, it is a very quick but informative overview of what a distributed database is. Key terms are identified and defined. Key concepts are listed and described. A very nice overview of advantages and disadvantages is also given.

Conclusion:

I did not learn a great deal from the article. However, I feel that the most useful part of the overview was the pro-con comparison of using distributed databases.

Top

Title:

11/8 - Summary of Embedded Systems, Linux, and the Future by Sebastian Niezgoda (hosted by: Karen Cosgrove)

Target Audience:

Linux developers/companies looking for embedded OS solutions.

Type of Document:

An overview of how Linux fits into the embedded OS market.

Summary:

The article provides a brief overview of the history of embedded system OS implementations. A number of companies are mentioned that have been leading the efforts in that area (Microsoft has not taken over yet). The author quickly moves to discuss implementation of Linux as an embedded system. Overall, such attempts are not shown as successful or even smart ideas. Linux is not a good OS for embedded systems, according to the article, for the following reasons:

• it's not a Real-time Operating System
• lack of standardization
• support of "industrial networking protocols"
• primitive USB device support
• lack of a standard GUI

Conclusion:

The author overall concludes that Linux is not the best choice for an embedded system OS. However, the developers should take it upon themselves to lead new efforts in that area.

Top

Title:

11/15 - Summary of Web Services Security, Part 1 by Sebastian Niezgoda (hosted by: Anthony Kim)

Target Audience:

A brief introduction into web services security issues.

Type of Document:

Article.

Summary:

The article quickly introduces the reader to web service applications, cross-enterprise/enterprise integration and service oriented architectures and related security issues. First, the security for integrated solutions is discussed. Network firewalls and SOA (Service Oriented Architecture) are provided as solutions to security problems; firewall for its use to filter, restrict, and encrypt traffic, etc., and SOA for using object oriented approaches, which allow for encapsulation and abstraction. Next, SOAP security is discussed. The article suggests that multiple servers are used to handle SOAP messages based on their security requirements or that SOAP packets are examined by the firewall. Finally, the use XML related security implementations are discussed. The article briefly talks about XML signatures, XML encryption, SAML (Security Assertion Markup Language), and XACML (eXtensible Access Control Markup Language).

Conclusion:

The article is a high level overview. It is not technical; rather it consists of quick and short introductions into the different solutions. I found it interesting because of its applicability to my class project - I have a better idea of which route to proceed (XML encryption and signatures) to implement security in my project.

Top

Title:

11/15 - Summary of Internet2 Middleware Initiative by Sebastian Niezgoda (hosted by: David Vaglia)

Target Audience:

Anybody interested in Internet2 initiatives.

Type of Document:

Website.

Summary:

The website explains that middleware is the "glue" that keeps the network and software applications working together.  Key projects are identified and sub-initiatives are described.  The site is a hub of information that anybody who wants to get a general overview of what role middleware plays in Internet2 efforts could consult.

Conclusion:

I would recommend the site to others because the Internet2 efforts are exciting.  It also features an introduction into middleware in general and how it will work on the new network.  The most important point highlighted by the consortium is that middleware should be a true infrastructure and should be used by all aspects of members (be it research labs or administrative offices).

Top

Title:

11/29 - Summary of .NET Distributed Transactions on Enterprise Services: a demo by Sebastian Niezgoda (hosted by: Abdo Achkar)

Target Audience:

Developers.

Type of Document:

Technical article and demo.

Summary:

The article walks the reader through a sample application, which transfers money from one bank to another.  Many code samples are included and explained.  Some different nuances are highlighted.  The reader is also shown how to run the example, which is written in VB but could be used on a .NET page.

Conclusion:

The article shows a very simple but interesting application of distributed systems code.  I highly recommend it to anybody who wants to quickly learn some distributed database coding (especially in .NET).

Top

Title:

11/29 - Summary of Initial End-to-End Performance Evaluation of 10-Gigabit Ethernet by Sebastian Niezgoda (hosted by:  David Acker)

Target Audience:

Network administrators, techies.

Type of Document:

An initial performance evaluation of an Intel 10-Gigabit NIC.

Summary:

The study examines the Intel PRO/10GbE LR Network Interface Card.  The card is one of very first capable of operating at 10-Gigabit network speeds.  The study evaluates the NIC on identical Dell servers coming up with initial end-to-end latency of 21 microseconds.  In a rather disappointed tone, the authors of the study explain their tinkering with MTU sizes, bandwidth-delays, etc.  In the end, the authors were able to narrow down the bottleneck to be software based, which was not satisfactory as the physical differences between the two test-beds were too varied.

Nevertheless, the disappointment of Intel's 10-Gigabit NIC was inferior to custom protocols.  The authors provided Myricom's Myrinet and Quadrics' QsNet running with 6 and 4.9 microsecond latencies respectively for comparison.

Conclusion:

The Intel card did not perform as well as expected especially considering near linear performance of the 1-Gigabyte chipsets.  Perhaps the problem, however, lies in the protocols and not in the interfaces.  Both Myrinet and QsNet reported around 30 microsecond latencies emulating TCP/IP layers.

An interesting read for any interested in implementing 10-Gigabit networks or any just interested in the direction Ethernet is heading.

Top