RobbieAllen.com > Blog

I’ve been very interested in Amazon’s cloud computing initiative. They’ve slowly been adding services to the point where some startups are able to use Amazon for storage, computing, and even database services. With StatSheet, I needed a simple content delivery network to host images, javascript, and stylesheets. S3 seemed like a good fit. I heard a few reports of S3 being unreliable when used as a CDN, but several bloggers I read seemed reasonably happy with them. I decided to take the plunge, change my some code, and start using S3 for static assets.

After a month I’ve decided to switch everything back to my own web servers. Why? Well of course there was the well publicized outage a couple of weeks ago, but the bigger problem for me is that periodically files would take an extremely long time to download. Occasionally I’d notice the site acting really sluggish. At least 3 times in the past month when that kind of slowness occurred, it was because an important file (like a javascript file needed for onLoad) is taking FOREVER to download. And by forever I mean on the order or 2-3 minutes! For just one file!

Below is a snap shot from Firebug that shows a 194KB javascript file being served from S3 taking 2.2 minutes to download! I could reload (with caching turned off) and it would take 2-3 minutes again. Each time I tried it.

Now eventually the file will start downloading fast again, but for an important file like that (or my main logo image), it dramatically alters the site’s appearance. Now I know there is a slight chance that it wasn’t S3 at all. Perhaps a network issue somewhere between me and S3. The catch is that all 3 times I’ve seen this issue happen, I can download the file with no problem directly from my web servers. I just don’t want to take the chance.

So I’m giving up on my S3 experiment for now. I’ll still use it for backups, perhaps, but as a CDN it just didn’t work well.

As I mentioned last year, I’ve stopped writing books and articles to pursue a new passion. Being the author or co-author for 10 books in 5 years can wear you out on writing ;-) However, an essay I wrote on blogging while at MIT has made it into a book that was just released. The title is Current Controversies: Blogs. The book is a collection of essays covering different aspects of blogging. I haven’t received my comp copy yet, but the table of contents looks pretty good.

For the first time in my life I’m starting to feel “old” :-) Having a child unlike anything else really makes you see how fast time flies.

But I wouldn’t trade this for anything:

That’s right. I’ve switched to a Mac. I bought a MacBook Pro a couple of years ago, but only used it sparingly. Now, I’ve turned in my Microsoft Club Card (btw, MS didn’t renew my MVP status this year and for good reason). There is no going back…at least for a while. There has been some initial discomfort during the move, but I’m getting used to OS X. The honeymoon period is already over though. I’ve experienced many application crashes. Firefox is still a pig. I’m going to give Safari a try. Because I’m still hooked on Outlook, I’ve been running VMWare Fusion. It is pretty good–especially Unity mode–but at times it slows down the system and doesn’t always suspend correctly.

Because I’ve been using Windows for so long as my primary desktop, it feels good to try something new. I have a theory that becoming too familiar with something is a bad thing. When you get too comfortable, you miss out on stuff. You never learn there is a better way to do something. Perhaps partly because I’ve written a bunch of books on Windows, I became too comfortable with it. There were no more surprises (besides the different ways it could crash on me).

Don’t get me wrong. I really really wish we didn’t even have to talk about which operating system I’m running. Modern desktop OSes are still too heavy handed. They should do all the heavy lifting behind the covers and get out of the way. I wish they were so dumb that no one needed to write books on them. But that’s not how it is today. Windows and OS X are big complicated pieces of software. With all the recent problems I’ve had with my desktop environment, I wonder if some higher power is pushing me to take on the cause of “thin” clients (for lack of a better term) because ultimately I still believe that’s where we are headed.

I’m both a BingoDisk (online storage) and Accelerator (virtual private host) customer with Joyent. I’ve only been marginally satisfied with BingoDisk to date (WebDAV access is flaky to say the least), and my Accelerators are decent but have experienced quite a few IO problems from being on shared storage.

As of last Saturday, BingoDisk and Strongspace have been offline. This wouldn’t normally be an issue for me because I access my BingoDisk infrequently, but it just so happens that my Lenovo T60p was starting to act up. I could sense a crash was coming. Unfortunately I couldn’t upload my latest files and pictures to BingoDisk (my online backup location). I figured they would get it back up any day. Well yesterday my system crashed and I spent a whole day trying to salvage the files I could. I lost data. Of course I’m not blaming Joyent for my laptop crashing, but it really sucks that when I needed BingoDisk the most it wasn’t there for me.

It looks like the problems aren’t just limited to BingoDisk and Strongspace. While working on another issue with Joyent, I stumbled upon this:

Ouch, that means their public site (at least one of them) is also down. The reason I went to their website is because one of my Accelerators has been experiencing IO problems all day. I have two Accelerators, one for dev and one for prod. The problem is with my prod Accelerator. Joyent support suggested I optimize my app. Eh. The problem sprang up out of no where last night. iostat shows an obvious bottleneck on the storage side. I asked if this was related to all the data being restored from BingoDisk and the support guy said it was unrelated. The only option they gave me (besides optimizing my app) was to move me to another host and take 2 hours of downtime. Ugh.

To summarize:
1) BingoDisk down for almost a week
2) Accelerator experiencing IO problems and only solution is to move me to another host.
3) Joyent’s own home page is down

Should I consider switching to another provider? Any recommendations?

Personally, 2007 will be a hard year to beat. I saw the birth of my daughter and have been enjoying her ever since. Professionally, it has been a mixed bag, with some things going well and other things not so well. Obviously the new baby had an impact on my priorities.

Below are my highlights (and lowlights) for 2007:

• My daughter was born :-)
• I got promoted at work
• For the first time in over 5 years, I did not write a new book
• I stopped editing books too! In fact, I don’t plan on doing anything related to writing or editing books for a while.
• I started a new pursuit
• For the first time in several years, I did not speak at or even attend a conference, go out of the country on business or pleasure, or play golf at least once a month.

Biggest highlight of the year? My pumpkin’s first Halloween:

About a month ago, I described how I was fed up with Firefox memory problems and instability so I decided to try out IE7. Well I’m unhappy to report that my experience with IE hasn’t been much better. Yes, it also bloats up to several hundred MBs and over a period of a few days my whole system starts to exhibit memory leak behavior. Part of the problem is that I’m running Windows Vista. I should have just stayed with XP. Perhaps I should “upgrade” back to XP. Or better yet, maybe I should put down my Lenovo and dust off my Mac.

It amazes me that in my 13 years of using the Windows operating system, stability is still a huge issue. Sure the UI has improved and the apps are more feature rich, but we’ve sacrified stability. I’m about fed up with it. The periodic need to reboot, memory leaks, SYSTEM process going to 100%, inability to handle multiple monitors well, video driver crashing after undocking, etc., etc.

All of this just furthers my belief that “thin web-based clients” are the future.

I stumbled upon an interesting post on Brazen Careerist titled “Twentysomething: Why I regret getting straight A’s in college”. I too was obsessed with getting straight A’s in college (undergraduate). I graduated with all A’s except 1 measly B. However, I don’t regret it. I don’t disagree with any of the main points made in the Brazen Careerist post, but I do disagree with the conclusions. Let me address each point and explain why.

1. No one has ever asked about my GPA.
Same here, but then again I didn’t do it so I could brag that I was a straight A student. In fact, I didn’t announce it to anyone. It shows up on my resume, but I don’t think it has done anything for me as far as being perceived as a “straight A student”.

2. I didn’t sleep
Me too! For a whole semester I only slept 3 hrs per day (using the Polyphasic method). But you know what? This instilled a great work ethic, which I carried over into my professional career (ie, cube sleeping). Not sleeping much doesn’t work for some people, but it does for me. Even now I need a max of 6 hours per night.

3. I’ve forgotten 95% of it.
Same for me. My brain is definitely wired to “use it or lose it.” That means I forget stuff relatively easily if I don’t recall it periodically. Most of the stuff I learned in college I haven’t had to recall since then, so it is lost in the nether regions of my brain. But is that really a reason not take classes (or read a book)? If you applied that same logic further, you should never take any classes (or read a book) because you’ll likely forget most of it. I believe learning (and brain development) is a cumulative process. The important point isn’t that I’ll remember some obscure detail, but connections are made in my brain that enable to me learn other things more quickly or to gain insights not possible without the prior learnings.

4. I didn’t have time for people.
Me either. But that has more to do with the fact that I’m by nature an introvert. Can you make a lot of friends and still make straight A’s? Absolutely! Study groups, clubs, intra-murals, etc. are all options. And you can still attend social events, but not every night. Would you want to do that anyway? Sounds like a pathway to alcoholism rather than making a bunch of friends.

5. Work experience is more valuable.
Work experience is critical. Your work experience will definitely provide more possibilities for career advancement than getting a 4.0 over a 3.0. That said, I don’t see the two as mutually exclusive. I did both. I had two internships at high tech companies while I was in college. I didn’t have to do any classwork during my internships and yet I still got some class credit. It was the best of both worlds and I highly recommend it to any college student.

What about Graduate School?
A big reason I wanted to do well in college was so I could attend MIT for grad school. It wasn’t until many years later that I realized that goal, but knowing what it took to push myself and do well in an academic setting was extremely helpful to perform well in a top school like MIT. (i’ll save the benefits of going to grad school for a later post.)

In summary, even with many years of experience under my belt and knowing what I know about the “real world”, I’d still recommend to an 18 year old Robbie to do his best in college. For type-A personalities like myself, it is very difficult to turn my passion for achievement off and on like the flip of a switch. I’m concerned that if 18 year old Robbie started intentionally slacking early in college, it would have continued into my professional career. I would have regretted not giving it my all.

I’ve been a long time Firefox fan. I’ve been using it since the early betas, but I’m so fed up with it now that I’m officially switching to IE 7 (and this despite the fact that I don’t like the IE 7 UI). The main reason for my dissatisfaction? 1) Memory consumption, 2) CPU utilization.

Microsoft gets a lot of flak for writing bloated software, but Firefox is the biggest consumer of memory on my system. A distant second is Outlook. It is not uncommon for Firefox to consume 500MB-700MB (Private Working Set) on my T60p (2GB) system. I may only have 4 or 5 Firefox windows open with 4 or 5 tabs open each, but Firefox will get so slow and unresponsive that I have to kill it. Fortunately, the “Restore Session” feature lets me restore my windows, but it won’t be too long before the memory usage goes back up and the CPU starts spiking.

Maybe I could blame this on my specific laptop, but I’ve had the exact same problems for a while now even on previous laptops. There has been quite a bit of press on the Firefox memory leaks and I know the Firefox team is working on it, but I just can’t deal with it anymore. Since much of my day is spent staring into a browser, I’d rather sacrifice a little usability (IE) than to sacrifice on performance (Firefox). Perhaps I’ll try it again after the next major release.

I’m a college basketball fanatic. I’ve been a North Carolina Tar Heels fan for as long as I can remember. Last year when I started thinking about interesting websites I could build, college basketball bubbled up as a potential topic. Currently, most of the big sports websites do a rather poor job of catering to fans that have a hunger for stats. Most existing sites use the same text-based, current season only stats as everyone else. We’re taking a fresh approach at StatSheet.com. Since I’m such a big college basketball fan (and blogger), I figured we’d develop a website that catered to my needs to start with.

We’re planning the first official release during Midnight Madness. This will include all the base features to view historical stats, compare players, coaches, teams, and conferences, graphical charts which can be embedded in your website, and much more. I’m announcing it now because we have enough completed to start getting feedback. Some pages will be rather empty, that’s ok, we’re still working on it. But we are interested in any general feedback about usability, functionality, and features you’d like to see added.

After the first release we plan a follow on release in time for the beginning of the season. This will include features that allow users of the site to contribute their insights and knowledge. That’s all I’ll say for now!

It has been a blast developing the site so far. Stay tuned for more!

The second edition of the successful Pragmatic Unit Testing book is heading to bookstores. This is the first paper book I edited for the Prags. Matt Hargett was responsible for updating the content in the second edition.

If you are interested in Unit Testing, especially using NUnit, this book will be extremely helpful. I’m not a C# person, but the discussion on Unit Testing practices along with the examples described in the book were informative.

One of the books I’m editing at the Pragmatic Bookshelf is Deploying Rails Applications. It just went beta this week so you can get your hands on an early version. It is a fantastic book and will prove very helpful to anyone that has to deploy Rails applications. I hadn’t made the leap to Capistrano yet, and this book guided the way. Good work Ezra and Bruce!

I’ve never been particularly fond of using titles to describe what I do or who I am. Since I work on a variety of things and have many different interests, it is difficult to label myself that way. However, on May 21, 2007, I earned two new titles that I am particularly proud of. The first is “Daddy”, which I think is a great way to describe my primary personal interest. The other is Distinguished IT Engineer (DITE), which I earned at Cisco. DITE is the top of the technical career path in Cisco IT.

I was the focus of a Cisco internal article about my promotion. I find it a little embarrassing, but my wife is making me post it here for my non-Cisco friends to see. As any smart husband does, I always listen to my wife ;-) I’ll post it as an image so the search engines don’t index the content. Click on the image to read it full-size.

On May 21, 2007 around 4pm, Landree Elizabeth Allen was born. She came just a little sooner than me and her mommy were expecting (by 5 days), but that was fine by us! It has been quite a whirlwind over the past 5 weeks (I can’t believe she is almost 6 weeks old!) We have enjoyed her immensely.

And I must say, she is the cutest little girl I’ve ever seen! And I mean that objectively! I promised myself that whenever I had a kid, I wouldn’t be like a lot of parents that tell their kids how great they are at something even if they stink. Or tell them how wonderful they are when they are really a little brat. Oh no, I wouldn’t be blinded by parenthood. That’s why I can say I’m being extremely objective when I say I have the cutest baby ;-)

UDPATE: A gentleman from the SBA wrote in to confirm the SBA never said that 90% of businesses fail in the first year. I asked for some referenceable data and he provided it. Thanks Mike! Here is a report from the SBA written back in 2002 that provides ample data that the success rate is much higher. I’ll jump straight to the punchline:

BITS shows that 66 percent of new employers survive two years or more, 50 percent survive four years or more, and 40 percent survive six years or more

The report also highlights how the SBA is frustrated by the 90% failure myth:

Although the business survival rates presented here simply confirm previous findings, perhaps this kind of independent confirmation is what is needed to dispel the myth that 9 out of 10 businesses close in their first year.

Next time you hear someone quote the 90% failure myth, correct them!

An often cited statistic (presumably from the Small Business Administration) is that 90% of businesses fail within the first 12-18 months.  This stat is typically used to illustrate just how difficult it is to be successful at starting a company.  I heard this many times while at graduate school including in my New Enterprises class where I was supposed to learn how to start a company.  It was very discouraging.

Fortunately, the stat is a load of crap.  Like many statistics that are thrown around in popular culture, a single percentage can rarely be applied to a large population.  Another great example of this is the divorce rate.  We often hear that more than 1 out of every 2 marriages end in divorce.  Those are some odds!  It really makes you feel good about entering a new marriage.  Fortunately, they are the wrong odds for most people.  The thing about the divorce rate is that it varies dramatically depending on your demographic:

if you are a reasonably well-educated person with a decent income, come from an intact family and are religious, and marry after age twentyfive without having a baby first, your chances of divorce are very low indeed.

The same applies to your odds of starting a successful company.  Even if 90% of companies fail in the first 18 months (which hasn’t been substantiated), that percentage would be applied across all companies.  The odds of someone with no training or financial backing being successful at starting a company is not the same as someone with a college education and a couple of businesses under their belt.  The demographic matters and there is empirical data to prove this.

When I was at MIT, I attended a lecture by Ed Roberts where he reviewed some of his findings from studying technology-based companies for 25 years.  Here are some relevant points from my notes:

• Who becomes technology-based entrepreneurs? (”becomes” not necessarily “who are successful”)
• Disproportionate share came from families where the father was self-employed
• Women were extraordinarily exceptional - very few
• Entrepreneurs have a moderate need for achievement, moderate need for power, but a high need for independence
• Development-oriented work background (not research)
• 50% of companies in his study were founded on a part-time basis despite this probably violating employee and IP agreements
• SBA says 80-90% of companies fail within 18 months, but their data showed only 16% failed after 5-7 years
• Average entrepreneur has less than a high school diploma, but their data showed avg tech entrepreneur has a college diploma
• Didn’t find a distinction based on where entrepreneur went to college
• 83% of entrepreneurs were highly satisfied with their performance 5-7 years post-founding and 17% were not satisfied. The entrepreneurs that were not satisfied did not correlate to those that failed
• Successful founders’ characteristics:
• Made, not born (found no birth characteristics that predict success)
• High need for achievement and optimally with moderate need for power
• If you have a high need for power you limit other people from stepping up
• Data is clear, individuals don’t succeed. The larger the founding team, the higher the chance of success
• Solo entrepreneurs seldom succeed, 2 are more likely to succeed, 3 are even more likely, 4 are even more likely, and 5 are even more likely (don’t have enough data after 5)
• Have a hypothesis that the more startups you found, the more successful they become
• Age distribution at first founding: gradual shifting to younger and older entrepreneurs (the middle is being vacated)
• Median age of founding first company is constantly decreasing: 1950s: 40.5, 1960s: 39, 1970s: 35, 1980s: 32, 1990s: 28

Ed Roberts data showed that only 16% of tech-based companies failed after 5-7 years.  Also, certain characteristics tend to predict success including larger founding teams and the number of companies started prior.

The moral of the story is that you shouldn’t believe every stat you hear.  On aggregate it may be correct, but when applied to your specific situation, your odds could be dramatically different (for better or worse).