Matt Ingenthron's Stream of Consciousness

Wednesday April 01, 2009

redirect

This blog is no longer active, so for further updates on my happenings, please have a look at blogs.ingenthron.org/matt

Monday December 22, 2008

Sun Web Stack 1.4 Release

Sun Web Stack 1.4 has now been released. You can download the release now if you plan to install on Solaris 10 or Red Hat Enterprise Linux. For OpenSolaris, the same software has been integrated into the package repository, so you can obtain the Sun Web Stack 1.4 release there by simply adding the packages you would like. There are some package clusters, like "amp" and "ruby" to simplify adding multiple packages, and there is a pretty slick, developer oriented GUI in the "webstack-ui" package. For more details on the packages in the OpenSolaris 2008.11 release, which aligns to Sun Web Stack 1.4, refer to the getting started guide.

I've been working with both Cool Stack and the OpenSolaris Web Stack projects for some time. Starting with OpenSolaris 2008.11 and Sun Web Stack 1.4 (the logical successor to Cool Stack 1.3.x), we've brought the two projects closer together. I'm sure this will raise some questions, and since I've been involved in both I think I'm in a position to help answer a few of them.

Why is the new release version "1.4"?

Astute observers have probably wondered why it's called Sun Web Stack 1.4 and not 1.0 or 2.0. Cool Stack users may recall that the last release was version 1.3.1. The choice of version 1.4 is a subtle way of saying this is the successor to Cool Stack 1.3.1. In fact, most of the components use the same optimizations (with a slightly different build process) and have been built and tested by the same engineers-- they're just updated components. The only exception is MySQL (more on that later).

Why did MySQL go from version 5.1 in Cool Stack 1.3.x to 5.0 in Sun Web Stack 1.4?

When we started work on Cool Stack 1.3, MySQL 5.1 was expected out any day, so it was updated to be current once Cool Stack 1.3 was released. Shanti and others in PAE did a lot of experimentation with MySQL 5.1 and found it to have a number of advantages. Problem was, the update didn't come out, and Cool Stack was ready to ship. We had to make a choice, so we decided to release Cool Stack with MySQL 5.1, with the thought to update it later once MySQL had a Community Server final release. The problem was, eventually we got to the point that Cool Stack was ready, but MySQL 5.1 was still undergoing development. So, release, or hold back?

Ultimately we decided to release. In retrospect, this may not have been the best thing since MySQL 5.1 was still evolving quickly. It's hard to say, since some people were using it. For others, it caused migration issues.

Regardless, for Sun Web Stack 1.4, we coordinated a bit further with our MySQL brethren and collectively decided the best thing was to ship MySQL 5.0. Yes, it's a version regression from what we had in Cool Stack 1.3.1. Cool Stack is still available though, and we don't have any plan to pull it. We'll also be looking to update things going into the next release. If you have any thoughts on this, it'd be great to hear feedback in the comments or on the Sun Web Stack forum

The paths seem different compared to Cool Stack... why were they changed?

For users who weren't using zones, the old approach worked. However, if you were using Solaris zones, or you decided to adopt Solaris Zones after installing into the global zone, then it didn't work so well. It was possible to just install into the non-global zones, but that did require some planning.

At the same time, I know there is a nice simple abstraction in having everything in a common directory. You look in /opt/coolstack and find it all there. The good news is that in the unbundled release, for the most part we kept this abstraction. The only difference is that this release uses /var/opt/ (variable files, like your htdocs) and /etc/opt (configuration files), which will work correctly with zones.

Having said that, I know that there are still those who prefer a zip or tarball style download which lives in a self-contained directory. There are some technical challenges in doing that correctly, but we are evaluating what it would take. If this is something you'd be interested in, we'd like to get some feedback.

Updates coming soon...

Tuesday July 01, 2008

Faban at LA JUG

Tonight (July 1, 2008) I will be speaking at the LA JUG on the Open Source project Faban. Faban is a benchmark harness/framework for developing and executing performance benchmarks against a variety of targets.

If you're interested in joining, please have a look at the LA JUG website.

Friday June 13, 2008

Another attempt with LA Metro I have a concert to go to tonight in the Echo Park area. To get there by 6:45 (1.5 hours from now), I'll need to leave an hour ago, take two buses and two trains. Oh yeah, and walk 3/4 of a mile. ( Jun 13 2008, 05:17:45 PM PDT ) Permalink

Sunday May 18, 2008

Why people don't take the Metro in LA

According to the LA Metro Website, with a combination of busses and trains, there is no way to get to Hollywood/Vine in 4 hours from El Segundo.

Somehow I think this is probably wrong, but making sense of it with this map is pretty much impossible.

Thursday November 08, 2007

The Canadian Double Double

I've spent a few days in Toronto with a customer working with Faban. Effectively, we took a generic benchmark I'd adapted from some work Akara in our pae group had done and helped them to use the Open Source Faban to test/validate what they're doing with databases on our systems with Solaris.

I'll post the sample later-- it should be pretty usable for anyone who wants to run through a set of operations against a database and see how they do. In the sample, I had implemented backends for PostgreSQL and MySQL so one could see how the operations fared on each of the different databases.

One of the little cultural differences I came upon here was that they have their own "Double Double". Here, that means double sugar and double cream in your coffee at Tim Hortons. It seems to be a variation on a Starbucks... kind of. In downtown Toronto, the Starbucks and Tim Horton's were all in close proximity, so I have no doubt they see each other as competitors. To me, culturally, it's closer to Dunkin' Donuts in Boston.

For those who don't know, Southern California's "Double Double" is a double meat, double cheese cheeseburger at In-n-Out Burger.

My friend here at Sun, Kim LiChong, tells me that part of the draw, or at least the branding, is that Tim Horton was a Hockey player for the Toronto Maple Leafs.

There were a couple of other cultural differences that I encountered-- the one other big one being that they actually changed the voiceover on the Mazda commercial to have a different pronunciation! Up here, like in Australia, it's "Maaaasda" as in the 'a' sound in apple, not "Mahsda". I saw the commercial several times-- I wasn't sure after the first time, but I was certain of it by the third time.

Thursday October 11, 2007

CEC 2007 is a wrap

Well, I got into and out of Las Vegas without doing any gambling. I did, in fact, attempt to leave some coins behind. However, I guess coins have no value to one-armed-bandits any more, so I left with my pocket change.

I did a little bit of talking, sharing what knowledge I have in one session, and did a lot of listening and learning. Some things like DTrace in Solaris, JavaFX and JRuby support in NetBeans have added functionality since they were first introduced. It was good to hear about these updates, and there is just so much going on at Sun it's hard to keep up with it all!

Of course, one highlight was the launch of the UltraSPARC CoolThreads T2 processor based systems. I've only been able to play with one of the systems so far, and even then not for very long, but I was impressed. The benchmarks show it, it's CMT again proving it's the right approach.

Scott Mattoon and I shot a brief video for CEC, which Scott posted on his blog here. ( Oct 11 2007, 10:57:31 PM PDT ) Permalink

Friday August 17, 2007

Another problem with raster graphics

Now, don't misinterpret the title-- I have no issue with raster graphics themselves when used correctly. The issue I have is many people unknowingly use them incorrectly and leave it up to the poor interpolation in browsers/mail clients to try to do the right thing. They're trying to do a bunch of things at once, so you end up with corporate logos that look all jagged.

There was a time back in 2001 or so when I was looking forward to SVG and others delivering the vector goodness that we should have had back in the mid 90s (NeWS anyone?). Sadly, it doesn't seem to have come to pass-- with the exception of plugins like Java with JavaFX Script and Flash.

Maybe it's a fault of mine, but this makes me think of whomever is using rasters in this way as an amateur. It immediately lowers their credibility in my eyes. I probably shouldn't think this way, but I guess because I spent some time massaging pixels and Bézier curves, that sort of thing bothers me....

Kind of like shoddy software on cell phones bothers me.... but that's a rant for another day....

Monday July 30, 2007

Transformers missed the "transform"

There are plenty of legendary (among geeks) Hollywood slipups with technical terms. This last weekend I encountered one more.

In Transformers, there is a scene where the blonde-model-playing-white-hat-hacker tries to impress upon the Secretary of Defense how sophisticated the inbound hack was. To underscore her point, she says it's not just simple "Fourier transfers" they're dealing with. The proper term of course is Fourier transform.

Ironic, given the name of the film.

Transformers Hollywood Fourier transform

Thursday May 10, 2007

What to Wear? The Modern Dress Code Problem

The other night, at JavaOne, I met Ben Rockwood, and as he and I were talking, Josh Berkus walked by. Ben complimented Josh on his hat, and Josh repaid Ben with a compliment on his kilt. I'm quite serious.

This is the manner of dress of most of my contemporaries and customers.

I, on the other hand, was in nice black pants, and a wrinkle free shirt (I was prepared for a customer briefing that didn't materialize).

Ironically, dress code was also a side discussion during the morning keynote. Scott McNealy (jeans, Sun logo'd shirt, $7 haircut) quipped about Jonathan Schwartz's dress (sport coat, shirt, tie, slacks, pony tail) as CEO, then turned his attention to Rich Green (nearly always in a black shirt, short sleeved) both playing the role of, and looking like Sun's Steve Jobs, though with short sleeves.

Over the last couple of years, I've started to dress a bit nicer. It was mostly due to the customer set I was working with- typically larger companies, but it was also because some Sun folks come into the office well dressed, as do customers and partners.

I guess the question is, does dress code matter? This question cuts both ways: I could be (and believe I was) perceived as not being someone capable of addressing technical topics, or I could be perceived as not taking business seriously.

Maybe I just shouldn't worry about it....

Tuesday February 13, 2007

Complete Open Source Stack at SCALE5x

Some time back, those who organize SCALE, a Southern California Open Source Expo, checked in with a couple of us at Sun to see if we'd like to be involved.  After all, Sun's pretty heavily involved and committed to Open Source these days.  Through the Los Angeles Unix community, I knew a number of the attendees as well.  

Since Sun is, to my knowledge, the company with the most complete Open Source stack, I thought it made sense at an Open Source expo to talk about that.  I figured most of the community knew about OpenSolaris (even though they were using other Unix-like operating system kernels with their GNU userspace).  Most of the community would not, unless you track Sun pretty closely, know about many of the other big projects: OpenSPARC, OpenJDK, CoolStack (released today as Solaris + AMP), PostgreSQL, Netbeans, project Glassfish and project jMaki.  So, I submitted a talk, which was accepted:

A Complete Open Source Stack: Hardware to Web 2.0*

That afforded me an opportunity to spread the word, show off the system, show off the software features like DTrace and ZFS.  It also gave me plenty of opportunity to talk about what Sun is doing that's different than the other commercial Unix and unix-like distros that run on x64 and SPARC.

As things came together, I found that Josh Berkus, fellow Sun Employee who is well known in the PostgreSQL community specifically and the Open Source community in general, was going to be at the PostgreSQL part of the show.  I invited Josh to be part of my session, and he accepted.  I think this added another whole aspect to the talk.

My talk, like many of them, was very well attended.  The room was full, and there were people standing in the back of the room.

Another advantage of these community events is networking and learning what others are working on.  There were definitely some surprises, and clearly some people are pushing the limits of the technology.  I had a few good conversations with people about Xen and how they're using it, and a number of good conversations about how people are putting together these modern web tier applicaitons in a virtualized space.

On the last few days, in the last few minutes, we had the opportunity to put together an OpenSolaris BOF session.  That did come together, and I gave a few updates on the OpenSolaris happenings.  We also decided in that meeting to create at least a virtual OpenSolaris user group of some sort (perhaps a UUASC associated SIG) here in Southern CA.

It was definitely a worthwhile experience and I want to thank the SCALE folks for having us.  I just got an email saying they had record attendance, so I'd expect that means it will be coming again next year. 

There was only one unfortunate thing.  It only came up twice, but some people looked at the "L" in the title and immediately assumed that Sun had no interest in a Linux expo. 

I have two responses for that.  First, it's very hard to say Sun has not made contributions to Linux.  It's probably true that Sun doesn't employ full-time kernel hackers, but Sun has contributed drivers, interoperability fixes and kernel fixes.  In my tenure at Sun, I've seen this done more than once.  Even then, Linux is more than a kernel.  I don't think it's possible to have a regional meetup of this size that would have only kernel hackers.  Sun's made numerous contributions to Open Source which are part of all of the modern distros.

Secondly, the SCALE event, despite the L, is not just about Linux.  Have a look at their about page.  It's an event for building the community of Open Source. 

In the end, pretty much everyone I talked to was happy to see Sun there and is happy to hear of Sun's work in Open Source. 

I was, in good spirits of course, accosted by a gentleman from IBM at the Sun booth about this very topic.  His question was what Sun is doing there at a Linux conference.  I reminded him of the above, and then proceeded to ask when components of Websphere will be Open Sourced, or when IBM will Open Source Power.

Admittedly, he and I are two leaf nodes in our respective organizations, so I'm sure he couldn't speak for IBM any more than I can speak for all of Sun, but the answers were telling.  He said he didn't see any need to Open Source Websphere.  In his estimation, no one is looking for the source.  He also said there was no need to Open Source Power 5 or Power 6.  In his estimation, if someone wants the specifications to the Power architecture, the community at Power.org has that for a nominal fee.

Of course, those in the know would be aware that IBM did apply an Open Source banner to a small part of the Websphere universe by acquiring a company and rebranding it Websphere Application Server Community Edition (even though there are major differences in the codebase between that and the WAS they'll sell you).  Those in the know would also recognize that what is at Power.org is much more like what's at sparcinternational.com and has been for over a decade.  Joining power.org and paying the (higher) fees to read the specs is more like being given the man pages with the opportunity to go write the code that implements the man page.  OpenSPARC is more like being given the code, the build scripts and the unit tests to go extend the most unique thread-rich chip anyone's been given access to yet.

This does seem to show, once again, that people get open standards and Open Source confused.  Adopters of OpenSPARC get both-- adopters of the power architecture get only one.

* By the way, I know that some take issue with the term Web 2.0, and I think I probably agree with the arguments against.  Having said that, I think we all admit it draws a crowd.
 

Monday December 04, 2006

Contemporary web architectures and Cool Stack

Since Marc outted the work I've been doing with our PAE and MDE folks, I guess I'd better type up some thoughts on the goings-on. It's time to get some of it out in the wild afterall.

Steve's new group, which I'm happy to be a part of, is off focused on helping Sun's Web Infrastructure customers. What does that mean? As many people have no doubt noticed, there's a new up-and-coming bevy of companies building out new products. Many of these are very consumer and social networking oriented, and others are just new takes on old ideas in a way that makes things more interactive and useful for the consumer of the product and/or service. Most readers of this blog probably are familiar with all of that and associate it with that currently buzz-laden term Web 2.0.

Brief sidebar: I've now been involved in and/or seen a few conferences where people have gone off and attempted to define Web 3.0. As an industry, we sometimes seem to be way more interested in going out and defining new things instead of completing or building out what we've just defined...

Sun's been working with many of these companies long term and has also been working with various startups as they come up. In the process, we've seen there are a few patterns that come up again and again. Sometimes, this is as straightforward as many people using memcached when scaling their PHP environment, but the pattern emerges when you start to see the Java world using things like JavaSpaces and Tangosol in their web application architectures.

So our goal? To help these people get the most of their web infrastructure, be it through scaling, consolidating or just applying a common pattern toward achieving their goals.

Okay, that sounds pretty lofty, so how are we doing that today? Well, we have a couple of compelling things in Solaris and some compelling things in our hardware. Specifically, if you could, for instance, have many, many threads running on the same cache of data with a T1000/T2000 at the same time you're reducing your datacenter footprint and power consumption dramatically, that would bring a benefit to some of these architectures. Another way to bring a benefit to these architectures could be to co-locate inexpensive, low power consuming, highly-reliable (through software) storage with core components of your application with the x4500. Therefore, we've been working with these customers on applying those patterns with whatever hardware/software is appropriate for the environment. Sometimes it'll be memcached, sometimes it will be lighttpd. Other times it will be the AMP stack, in this case embodied through Cool Stack.

Speaking of Cool Stack, the deployment oriented, performance optimized AMP stack for Solaris, there are some good things coming and we'd be interested in what you're interested in. Please either post a comment, post a blog with a trackback, or drop me an email. Of course, the latter of the three is a decidedly Web 1.0 approach. :)

I will be blogging more on the specifics in the not-too-distant future.

Technorati tags: memcached, lighttpd, Web 2.0, ZFS, OpenSolaris

Tuesday July 18, 2006

Trackback problems due to failed verification with JRoller?

Okay, I admit I don't always read the directions. Lately though, when working with my blog, attempts at posting a trackback would give me something like this:

<?xml version="1.0" encoding="iso-8859-1"?>
<response>
<error>1</error>
<message>ERROR: REJECTED: trackback failed verification</message>
</response>

Target blog returned HTTP response code 200 

I'd looked into the problem once, and John Clingan and I even did some basic testing to see why, but couldn't make sense of it. So tonight, I tried it again with the Scott Jolly post, and had a similar issue. After a minute or two of research, I came across a note in an email from Allen:
trackback verification ensures that the entry you are sending the trackback from has an actual link to the entry you are sending the trackback to. So if you want to send a trackback from entry1 to entry2 then entry1 must have a link to entry2

Ah ha! Well, sorta, I had a link to Scott's blog already. Hmmmmm.... Well, let me give it another shot-- I'd not posted the blog entry yet when I tried to set the track back I don't think. Hey, it worked!

Dave, if you're listening, that trackback verification error message could be a bit more specific about what a verification is.... :)

Scott's Challenge in working from home

Woah Scott, since you've been blogging on working from home, let me suggest looking in on the kids every once in a while. :)

Once, when my parents weren't looking, my brother and I discovered that charcoal briquettes could be used to write on the outside walls of the house. We had redecorated the whole back yard!

Monday July 17, 2006

Back from the Sun holiday

It's been a busy few weeks, so just a couple of updates...

Sun Java System Portal Server seems to be hot these days. I've done a few demos and proof of concepts in recent weeks. They've all gone pretty well I believe-- Portal 7 has definitely added some needed features and has done much to get close to the world of composite applications.

Things should get even more interesting, because at Java One, Sun announced the next-generation portal server will be coming out as Open Source. Some of this has already happened, as you can see here, here and here. I've joined both of these projects and already thought about some places I'd like to make some changes. Now I just need to get the time. :) Don't be surprised if a new category comes up on this blog soon.

Welcome to the blogosphere Tom, Paul, Karthik, Manish, Deepak, Unnikrishnan and Hemadri. Check out Tom's first entry for a history of where we've been.

Other interesting things: a trip to the bay area (Marin county) to hang out with some friends, and a quick trip to San Diego. There were a few concerts in there as well, the San Diego trip was to see Radiohead not once, but twice. Then the day after Radiohead was Beck.

Radiohead is testing the waters with some new songs, but otherwise the concert was the tried-n-true standards. No major surprises, and a worthwhile concert. The venue in San Diego was good, but I must be an old dude or something because I can't stand all the shoving and cutting in front of people with General Admission concerts in a field. For the 2nd concert, because the venue was so good, we sat in the seats they had in back of the crowd. The sound was slightly better up front, but it was good in both places. I think I do like what will ultimately be the big new song Down is the New Up, despite not being sure what the lyrics are yet. :)

Beck was an excellent (as usual) concert. Beck was probably my favorite concert last year, and it's looking like it'll be my favorite concert this year.

We didn't get to my favorite Brazilian barbeque restaurant in San Diego though, Rei do Gado

Technorati tags: Portal, Bloggers