New Sun SSDs Could Spell The End of Root Disk Mirroring

Solid State Drives are on the way to Sun.  Some of us in mid-range computing have been dying for this development to come to prime time and frankly I’m excited.  Currently, most Solaris servers are using Solaris Disksuite or Veritas in order to mirror the root drive.  This provides the ability for the computer to still boot and return to service, even if one of the hard drives has failed.

But if you had a SSD for a rootdisk, it would consume about 25% of the power that the current drives do.  Also, they’re much faster, cutting boot time in half.  And ultimately the most important feature is that they last 7 times longer between failures.  Statistics were taken from this post.

Also, SSDs tend to not actually actually catastrophically fail instead they begin to throw more and more read errors.  This error rate would be accommodated for by any modern filesystem and would result in a minor slow down and an error being thrown to the kernel.  After enough gradual degradation, I’m sure Sun would agree that the drive is toast and then agree to replace it.

But think about this for a moment, it’s entirely possible that SSDs would become the defacto platinum spark plug for servers.  If a Solid State rootdisk lasts seven times longer than current rootdisks, it’s entirely possible that most Sun servers will meet their End Of Life schedule before the hard drive actually even fails.

I’d love to talk to anyone who’s had some hands on time with some of this hardware.

Perl is going away? No way Jose.

Unbeknown to me, some people have been asserting that Perl’s heyday has come and gone.  Sterling Hanenkamp and Kevin Marshall have posted a fair rebuttal over at perlbuzz.com.  I thought I’d tag on a bit to what Kevin and Sterling brought up.

Truly great languages do receive updates less frequently as they age.  That’s because such a large base of programmers have learned the earlier libraries and being the bright chaps they are, they requested more and more features.  Until at a certain point the product becomes so eminently usable that very little else needs to be added for most users.

Now there are going to be specific requirements for each project and that’s where the development comes in.  But god bless these bright chaps, because not only do many Perl programmers write this custom code, but they frequently check it back into CPAN as modules.

So other people who program similar projects with these precise requirements don’t have to reinvent the 7-spoke, 16″, lithium-greased, carbon steel wheel.  You see, Perl provided the wheel to everyone who worked with the base language.  But not everyone needs a 7-spoke, 16″, lithium-greased, carbon steel wheel.  Especially when they’re writing really secure code that has to be analyzed line by line for security risks.  Or if they want to write really small code that will work on a mobile phone or embedded code.

Why is this good for business?

Do you really want your programmers to have to re-learn their code-base every 5 years?  I knew lots of people who used to know how to do some Visual C++ and Visual Basic development back in the .com days.  Then along came Java.  In the midst of this came a lot of middleware that provided common libraries for programmers.  And then came .net.  Now, most Java developers don’t do Microsoft stuff, but these days some .net guys do perl and some java guys do some .net.

But trying to find and hire people that will mesh with whatever mix of code your shop is currently running is difficult for technical managers.  It’s completely impossible for HR and non-technical managers.

Now I’m not saying that hiring Perl developers is necessarily any easier.  And I’m not trying to be a Luddite either.  All I’m saying is that new revisions of products come out when current needs aren’t being met.  So we haven’t needed another Perl for a while now.

And Sterling hits on a brilliant point about how poorly the TIOBE index measures current market needs or the current state of code development.  Go figure that a free language doesn’t have a bunch of PR geeks hiring Search Engine Optimization guys to drive up the results of people’s searches.  Microsoft and Sun are literally spending millions of dollars to try in tilt these sort of comparison’s in their favor because it helps them make money.  Then we have to spend our money to learn the new skills, license the new libraries, call for support on their poorly written new projects, etc. etc.

In summary, Perl still kicks much ass and it’s not going anywhere any time soon because the price is right, it does the heavy lifting, has a good community full of nice people like the Perl Monks and Larry Wall is god.

A designer knows he has achieved perfection not when there is nothing left to add, but when there is nothing left to take away.
Antoine de Saint-Exupery

homemade electric motorcycle does 300MPG

but it’s completely unfeasible for the auto industry to possibly turn out a four wheeled version in under a century. Oh wait, they did, and then they recalled them all and crushed them. Watch Who Killed the Electric Car? for more on that story.

Back to the motorycle: here it is in all it’s splendor: http://feeds.gawker.com/~r/gizmodo/full/~3/265714089/homemade-electric-motocycle-gets-300mpg

The range is much too short for my needs.  But if I had a plugin hybrid with a 30 mile range, I would be perfectly content.  Slap some photovoltaics on the roof and then I’d really be chugging along for cheap.

Advanced Sun LDAP administration

Today is busy as it’s my first day back in the office after 4 days at class. It was a good class on Advanced Sun LDAP administration. Great guy Kevin from Chalk IT in Canada taught it. Nice guy, but don’t ask him about his speeding ticket. Fellow won’t shut up.

Anyhow, I’m off to get coffee in order to gird myself for the upcoming meeting at 9. Then I get my last year’s review at 1. How fun.

Sun iPlanet Replication fixed

So we have escalated on Sun enough that they brought in an expert, Fred Munday.  He reviewed the work, our procedures, and they even tried performing the event for us once.  It turns out we have one of the largest iPlanet LDAP systems in the world.

On an impulse, Fred just took the binary file itself from the original master LDAP, moved it over, and started the backup master LDAP process.  Theoretically this shouldn’t work.  But it did!

I’m actually writing this post years after this, so the date of the fix is an estimate.  If I recall correctly, one of engineers also realized that the file transfer between the two boxes was taking too long.  He figured out that the routing was set up wrong and it was going over a Gigabit NIC configured to run at Fast Ethernet(10% of it’s function) speeds.  So he set up the copy to run over another Gigabit NIC and that really increased the speed of the transfer.  Not by 900%, but much faster.

Sun iPlanet Replication still down

event didn’t succeed. AGAIN! I’ve been doing this same procedure for about 10 days now. It’s never worked. Not once!  We’re trying to get a multi-master replication in place where data is synchronized in both the LDAP masters.  So you export the data to a LDIF file, copy it over to the backup master and import it.

Unfortunately it takes close to 24 hours to perform the export, and 20 some-odd hours to transfer it to the other system and then more than 24 hours to import it.  Then the changes that have occurred in the last 68 hours on the original DB have to be synced up to the backup master.  But we only retain the last 72 hours of changes and it takes more than 4 hours for the final synchronization to catch up.  So some of the changes never make it over and you have an imperfect mirror.

Additionally the migration is now approaching. Which means more sleepless nights! I pretty much need to just understand that I’m going to be in and out of the office for a few hours at a time for the next few months. I need to give up and just forward my desk phone to my mobile and put an out of office sticker on my monitor or something.