Esoteric Curio - Writing

Scalability in Curriculum

nospam@example.com (Theo Schlossnagle) — Mon, 27 Oct 2008 23:50:44 -0400

I'm exciting that the Universidade Federal de Goiás or Brazil has a course called: "Scalable Architectures for Large Scale Applications" in their Computer Science department.

What's more? I'm very honored to learn that they are using my book as a part of their curriculum. I've added a venue for recruiting Site Reliability Engineers at OmniTI.

Why I didn't write for Wrox.

nospam@example.com (Theo Schlossnagle) — Tue, 16 May 2006 21:13:07 -0400

I've been asked by a lot of people why I wrote for Sams. Actually, I've only ever been asked that once, but I've been asked countless times why I didn't write for publisher XYZ (O'Reilly, APress, Wrox, etc.). Well, I suppose I found the best reason of all. Today George threw together a possible cover for my book in classic "wrox-style." I liked it so much, I pulled touched it up in Illustrator and have a nice shiny PDF. How many people do you think would buy my book if Wrox published it?

Watch out, I will eat your head.

The End Has Come

nospam@example.com (Theo Schlossnagle) — Thu, 11 May 2006 07:41:07 -0400

For those of you who know me, have seen me present or ever actually read all the various bits of my blog, you know that I've claimed to have been writing a book. In fact, it's been listed on Amazon for about two years. People have back ordered it there, and, I'm sure, feel like they will never get it. Well, the end has come. I have submitted all my materials to my publisher and it should be a matter of weeks before it goes to press.

The entire writing process was not an easy one. It was a challenge to present both high-level information as well as low-level implementation details in a fashion that focused on architectural design rather than specific technologies. After reading other books, I always felt that the authors were attempting to demonstrate how to solve a specific problem by scalaing a specific vendor's product. The truth is that scalable design is a frame of mind. It is about using an understanding of the points of contention and where pressures will sit on various components and applying good engineering practices to shift (or eliminate) those stresses. By presenting "how to solve problem Y using product X," readers are left in a bad position. While it might be easy to acquire product X, it is impossible to simply make your problem into problem Y. You have your own problems and you need the mentality to solve them adequately -- not a canned solution to a different problem. Can you say: "square peg, round hole?"

The book is called Scalable Internet Architectures and it discusses a variety of issues related to the design and managment of really large systems. In a two dimensional world, performance and scalability are orthogonal axis and the goal is to get as far from the origin as possible. If a lot of effort is put into the engineering of an application, then performance can be increased -- but it is bounded. If the right engineering effort is put into scalable design, the gains are boundless. My goal is to teach scalable systems design through conversation and demonstration.

The book is a good mix of high-level principles for managing big systems that walks you through best practices, technological terms, and concepts as well as extremely low-level implementations that solve scalability issues.

I had a lot of fun writing the book and I hope you have a lot of fun reading it (and unlike my blog postings, it was professionally edited)! Once the book is actually on-the-shelves, I'll see if I can get a release to put a PDF of the "Introduction" up here.

Apache 2002 US slides (a little late)

nospam@example.com (Theo Schlossnagle) — Wed, 07 Sep 2005 08:31:38 -0400

Today someone posted to the Wackamole list and asked for some detailed documentation of the configuration options and other such things. I looked around for a bit and there is very little. However, I knew that I had written some.

After some digging (and it took a while), I arrived at my ApacheCon 2002 US tutorial presentation. Amazing when you get paid to do something it is usually quite thorough. Anyway, 3 years later... here are is the tutorial slide stack.

Understanding and building HA/LB Clusters
(slide stack)

OSCON database replication slides.

nospam@example.com (Theo Schlossnagle) — Fri, 05 Aug 2005 10:00:00 -0400

It was a pleasure speaking at OSCON2005 on Wednesday on the topic of cross-vendor database replication approaches. Many attendees requested that I put the slides online. While there is a tremendous amount of commentary and critique that goes with the slides, I hope that they will be somewhat useful to a more general audience. Alas, here they are:

ApacheCon 2005 Scalable slides.

nospam@example.com (Theo Schlossnagle) — Fri, 22 Jul 2005 12:00:00 -0400

For those that are interested, here is a PDF (print form) copy of my Scalable Internet Architectures talk I gave at ApacheCon 2005 in Stuttgard .

Scalable Internet Architectures

nospam@example.com (Theo Schlossnagle) — Sun, 11 Jul 2004 00:03:24 -0400

I've wanted to write a book for a while -- mainly because people seemed genuinely interested in the presentations I have given over the past several years at the ApacheCon conferences.

While my academic background and most of my project background is in the area of high-availability and resource allocation in clusters (i.e. load-balancing), it seemed that a book on it would either be too high level or too applied. The problem with high level conversation is that you end up boring people to death who are practical and not providing enough hands-on information to those that are engineers. Too low-level is usually not a problem in the field of technology, but HA/LB systems are traditionally commercial and a hands on book starts to look a lot like a product's user manual.

Where to go from here? Well, I am not a big fan of a book on one piece of software, specifically when it is a low level systems tool. Books on Oracle or MySQL or Apache are quite useful if you want to learn about the specific software products. The problem with HA/LB is that it is a small layer on larger an architectural issue that is deep and all-encompassing. That issue is scalability. Concentrating on one technology in particular tends, it is easy to deviate from the purpose -- the purpose being "the methodology of scaling systems."

Scalable Internet Architectures will be my stab at tackling various aspects of scalability in today's web architectures. The focus is on scalable system design methdodology and "thinking scalable." Topics include:

Managing large systems
High availability
Load balancing
Highly-distributed static content delivery
Dynamic caching technologies
Databases and Database Replication
Clustered Logging
Building highly customized tools to tackly acute performance problems

Additionally, I will talk (at various depths) about the following technologies (short list):

Apache, thttpd, Squid
Linux, FreeBSD
MySQL, Oracle, Postgres
Wackamole, CARP, VRRP
DNS, Anycast (shared-IP)
Spread
perl, PHP, mod_perl, Apache::ASP
RHT, Splash!, memcached, NBD
... loads more.

The goal of the book is to look at the typical stresses in a production architecture as the demands on that architecture increase and to walk through solutions and understand how they alleviate those stesses.

To appease myself, there will be a sufficient amount of theoretical talk (a.k.a. idealism without regard to practicality). By being throwing idealist perspectives in where appropriate, the hot burning flames of product and technology propaganda can be kept under control.

I hope that readers are able to gain two vital skills from reading the book:

the ability to look at systems and understand what the bottlenecks are/will be and apply basic techniques to increase horizontal scalability,
to take every new technology with a grain of salt and thoroughly understand its limitations and therefore better understand how it can (or can't) be placed to benefit an architecture.

... wish me luck.