Performance Dynamics Consulting
Castro Valley, California, USA
| Where? | When? | What? | How? |
| UCLA Extension | March 18-22, 2002 | Performance analysis | Register online |
| Room G-33 West | Monday-Friday | Scalability modeling | at the UCLA web site |
| 10995 Le Conte Ave. | 9am-5pm | 3.0 CEU credits | Use number: M4429U |
UCLA Extension is adjacent to the
Medical Center
and UCLA campus in Los Angeles, California.
Both
Course Content
and
Registration
are available via the UCLA website, or
call the Program Office at (310) 825-3344 for more information.
Criticism of Gnutella network scalability has rested on the bandwidth attributes of the original interconnection topology: a Cayley tree. Trees, in general, are known to have lower aggregate bandwidth than higher dimensional topologies e.g., hypercubes, meshes and tori. Gnutella was intended to support thousands to millions of peers. Studies of interconnection topologies in the literature, however, have focused on hardware implementations which are limited by cost to a few thousand nodes. Since the Gnutella network is virtual, hyper-topologies are relatively unfettered by such constraints. We will analyze the comparative performance of several plausible hyper-topologies and compare their throughput up to millions of peers.
I will show you how to build a model of a threaded server in PDQ.
This relies on constructing what is known (in the business) as a Flow-Equivalent sub-model.
This is necessary when you are trying to determine how many threads should be available to service an application.
A comparison with commercial web application servers, such as Web Logic, will also be given.
There are some tricks to capturing the concept of parallelism in PDQ.
I will demonstrate how by examining some examples in detail.
We will take a detailed look at how to model multiprocessor servers using the following queueing algorithms:
- Erlang A (Kidding ... There isn't any Erlang A 1)
- Erlang B
- Erlang C
- Approximate Erlang C
- The Repairman model (singular)
- The Repairmen model (plural)
Models from Chapter 7 of The Practical Performance Analyst textbook on Multiprocessor Systems will be discussed in detail.
The plan is to spend time late Thursday and early Friday going through the process of model construction in PDQ.
The software itself can be downloaded at anytime.
Since this is open source, you do not need to have purchased a copy of the textbook to use PDQ.
To participate effectively in the UCLA class, you should bring a laptop computer with your favorite C compiler installed.
Better yet, download and compile the PDQ source prior to attending the class!
1 But do you know why not? Come and find out.