I recently reloaded mine with Windows 7, which was ridiculously easy.  Windows 7 runs no slower on it, and perhaps a bit faster.  Windows 7 natively loads all required drivers, so you won’t spend hours looking for the NIC driver or whatever for this model.  I like Windows 7.

Still as an older machine, it’s a bit slow.  The hard drive seems to thrash.  People say that replacing the internal hard drive with an inexpensive solid state drive (SSD) helps.  Will it make any real difference?  Yes.  I did it, and I’m glad I did.

Like I said this is HP Compaq tc4200 Tablet PC.  It has 2GB of RAM in it (the max).  It came with an 60GB 5400RPM ATA/IDE Hitachi Travelstar 5K100, model HTS541060G9AT00. It is loaded with a default load of Windows 7 Enterprise, fully patched.

Here’s some quickie benchmarks:

• 43 seconds to go from power on to login prompt
• 12 seconds from login to desktop
• 6 seconds to launch Word

I loaded iometer 2006.07.27 (a disk benchmark) on it, following my instructions.

Here’s what it looks like running, after about an hour:

iometer reports something like:

• 500 I/Os per Second
• 6 MBs per Second
• 20 Average I/O Response Time (ms)

I then replaced the Travelstar with a Transcend TS64GSSD25-M 2.5″ 64GB PATA MLC Internal Solid State Drive (SSD).  I got it for about $200 at Newegg.  I loaded Windows 7 and iometer on it again.  Yes, it is something like 50% faster on the simple things I timed.

• 27 seconds (was 43)  to go from power on to login prompt
• 5 seconds (was 12) from login to desktop
• 2 seconds (was 6) to launch Word

Note that benchmarking SSD’s is notoriously inaccurate, basically because SSD makers know you’re going to do it, and tweak the SSD’s so they behave very well in the first few minutes.  They know the average person will probably do some quickie benchmark on the SSD in the first few minutes of use, so they make sure the SSD is configured to respond quickly during initial use.

You can get past this by running the benchmark for a few hours, at least 8 hours or overnight.  If you watch the results over time, you’ll see the SSD performance drop by 50% or more from the initial numbers.  Of course the initial numbers are what the manufacturer puts in the specs and on the box.  Basically, they lie.  For example, Transcend says this device gets 50 MBs per second for writes.  Yeah, sure it does.

So here’s the iometer run against the Transcend SSD, after about 8 hours:

iometer reports the SSD is doing about:

• 2500 I/Os per Second (IDE was 500)
• 33 MBs per Second (IDE was 6)
• 4 Average I/O Response Time (ms) (IDE was 20)

So yes, it’s faster.  Maybe 5 times faster.

In usage, the tablet feels “snappier”.  I like it.

You’ll have to decide is $200 is worth making your old laptop faster.  If it’s some old beloved model that you can’t bear to part with, then maybe it is.

This test is long, longer than any other Microsoft test I have taken.  It has twice as many questions.  I took the test in Bakersfield, at Diamond Technologies (nice place, I’ve been there twice now, probably the nicest test center that is anywhere close to me).  The test proctor sat me down and said “You have 4 hours for this test, good luck.”  I about fell off the chair.  Four hours!  I finished in a little less than three, and that’s with taking the time to go over my answers a couple of times.  Previous exams were more like 90 minutes.

The test is much like the book, and a lot like the practice exams, unlike some of the other tests in this series.  Like I said before, the Microsoft 070-647 Windows Server 2008, Enterprise Administrator exam is only sort of like the book, and it doesn’t look much like the practice exams either.  I taken it twice, and have yet to pass it.

If you’re curious,  I keep all my certifications here.

Here’s a screenshot of it in action:

What was different?  I went slower.  Read the book again.  Took the practice tests over and over again until I was blue in the face.  Actually fiddled with some of the technology on my own.  Not much different, actually.

When you log in as root on a Nexenta box, you get the nmc.

People that really like to break things fiddle like an honest to goodness root shell.  If you really want a shell, then do:

[glporter@workstation ~]$ ssh root@my_filer
Password:
Last login: Wed Jun 2 11:34:10 2010 from workstation
nmc@my_filer:/$ option expert_mode = 1                                          

nmc@my_filer:/$ !bash                                                           
You are about to enter the Unix ("raw") shell
and execute low-level Unix command(s).
Warning: using low-level Unix commands is not
recommended! Execute?  (y/n) <press y>

root@my_filer:/volumes# <do special root-ish things here>
root@my_filer:/volumes# exit
exit

Important: To re-sync the appliance's management
state information, please consider running
'setup appliance nms restart' command.

nmc@my_filer:/$

Probably is a good idea to run ’setup appliance nms restart’, seeing as they told you to…

In general, file systems are not exactly glamorous.  HFS?  UFS? VxVM?  Who cares.  It’s a file system, right?

ZFS has some serious juicy goodness.  Personally, I think as time goes on, you’ll see more and more ZFS (or at least ZFS-like file systems) used on production servers, especially file servers.  Like NexentaStor.

One of the things I like best about ZFS is that the developers worked hard to make it easy to use.  ZFS is a combination of file system and volume manager.   What this means is that ZFS in dead simple to administer.  Most commands are one liners, and take the place of what would have been whole strings of commands with other volume managers.

Believe it or not, most file systems *DO NOT* actually detect or repair “bit rot”.  Hopefully you have underlying hardware like a RAID controller to handle that.  Even if you do, all it usually does is perhaps complain about “corruption detected”.  Bits flip and files get whacked more often than you think.  Not with ZFS.  ZFS uses elaborate checksumming to detect and repair problems.

Along similar lines, ZFS has atomic writes.  Either something is written to disk correctly, or it is not written at all.  You see high end databases like this, but not file systems.  What this means is that you can literally pull the plug on a server during a big write, and the file system stays intact.  No more fsck!

Here’s ten reasons to reformat your hard drives.

To summarize: Good stuff!  If you are looking for storage appliance software, try Nexenta first. Maybe you’ll luck out and Nexenta will like your hardware. If not, run Openfiler.

The Nexenta Community Edition is free to use. They do make you register it. That’s a bit of a hassle, especially if you fat-finger the key (”Is that an I or a 1?”)

The Nexenta load is no harder than, and arguably easier, than loading Openfiler.

Nexenta goes faster that Openfiler.  Nexenta goes faster than my $25,000 Dell/Equallogic PS5000E.

I’m working on a step by step how-to for loading Nexenta.

• Nexenta is more picky about hardware than Openfiler. Some machines that run Openfiler fine won’t run Nexenta at all. Since Nexenta is based on OpenSolaris, the OpenSolaris Hardware Compatibility List may give you some guidance whether your hardware will run Nexenta.
• Speaking of pickiness, I’ve had Nexenta repeatedly give me grief at install time on certain machines. The installer starts, doesn’t get very far, and then reboots. I found that Nexenta doesn’t like the BIOS APCI power settings. If you see this, try fiddling with APCI in BIOS (like turning it off).
• Openfiler is a bit “fiddly” at times. Something that you know for a fact works, doesn’t, and you wind up dinking with it. For example, I almost always wind up throwing my first iSCSI target away, and making a second one.
• If Nexenta loads at all (see comment about hardware pickiness above), it just works the first time, every time, smooth sailing.
• Openfiler 2.3 “out of the box” is broken. iSCSI luns will present once, but will not present again after a reboot. This is well discussed issue. It’s easy to fix, but it’s unfortunate that the Openfiler people haven’t addressed it.
• I like the built in patcher in the Openfiler GUI. I’m not sure how Nexenta does patching.
• Nexenta is prettier, especially if you have flash. The little tachometer gauges are especially cute.
• Some of the advanced features in the pay-for-commercial NexentaStor 3.0, like the High Availability Cluster Plugin are supposedly easy to deploy. Openfiler has features like this as well, but few mortal humans have ever figured out how to actually use them.

I read the book for this exam multiple times.  I painstakingly did all the labs.  This book does not cover test topics to the same depth and detail as you see them in the test.

I took the practice tests over and over again until I had all the questions memorized and repeatedly got perfect scores on them.  The actual test is unfortunately not a whole lot like the practice exams.

Oh well, Microsoft has their Second Shot program going right now, so retakes are free.  The program expires at the end of June 2010, so I will have to do the retakes soon.

So anyway, things seem more stable now, and hopefully no content was lost.

Please email me at the usual address (or use the handy “Email Greg Porter” tab on the upper right)  if you notice anything goofy.

Thanks!