MicroSD HD Tune benchmarks

I recently picked up a Sandisk 64 GB MicroSDXC card for my Atrix HD phone. The reason for the purchase is that I've always wanted my entire music library on my phone. I struggled through this over the years with various devices because my library is about 35GB. My first gen 16GB iPod Touch wasn't enough and then my third gen 32GB iPod Touch almost cut it. I got my first Android phone the Atrix 2 and wanted to get everything on it but 64GB cards were barely on the market and very expensive. So I was then limited to 8GB (32GB cards were still pricey too) and because of this had spent less time listening to stuff in the car. The Atrix HD came out this year and I purchased a 32GB MicroSD card hoping to get back into the tunes.

Google Music popped up and I decided to store everything there and stream to my phone. But then came the whole problem with having a data plan. 200MB a month wasn't going to work and I quickly found out that 300MB wouldn't work either even without streaming music (maybe a song or two) , just e-mail, navigation data and random browsing put me over the limit twice. I had to upgrade to a 3GB plan this month and having just signed a contract in July I couldn't go over to Sprint for unlimited data (and at the time none of their Android phones had a MicroSD slot). I'm finding in the last couple weeks that for the amount of music I listen to 3GB probably won't work either. So here comes the 64GB card and I can go back to 300MB to save some money.

Noticing the advertising on the box of the new Class 10 MicroSDXC card I wanted to benchmark with HD Tune to see how it compared to my old generic 32GB and a stock 2GB card that came with a Blackberry Bold 9650. I also used the BB to do some of the speed tests because the BB formatted card would not work in the Atrix.

Devices:
BB Bold 9650
Motorola Atrix HD MB886
C2D on Windows Vista x64
USB 2.0 connection

Cards:
SanDisk Ultra 64 GB microSDXC Class 10 UHS-1
RH Data 32GB HX Class 6
SD-C02G Taiwan MicroSD

Results

The short summary is that the 2GB and 32GB cards have basically the same performance, while the Sandisk nearly doubles the performance. But we do see a difference in speed with the same card between the two phones, so the phone performance does have noticeable effect. The odd thing is the high CPU usage with the 64GB card.

Cards	BB Bold 9650	Atrix HD MB886
2GB	Min: 7.7 Max:  12.4 Avg:  10.0 Access:  1.8 Burst: - CPU:  9.1%	*Card could not be read
32GB	Min:  8.5 Max:  10.6 Avg:  9.5 Access: 1.4 Burst: 7.4 CPU: 17.3%	Min:  10.7 Max:  17.2 Avg:  15.9 Access: 1.5 Burst: 23.1 CPU: 13.1%
64GB	*Card could not be read “Fatal error”	Min: 16.5 Max: 27.8 Avg: 26.6 Access: 1.2 Burst: 22.3 CPU: 22.2

Pictures

Bypassing the Start interface in Windows 8

*Note, 7/1/2013. The boot to Desktop mode has been included in Windows "Blue"8.1 (right click on Taskbar > properties > navigation>) so this post is no longer relevant. Also, I cannot stand Windows 8 (hah!)

I'm currently attempting a migration to Windows 8 at home and at work and have discovered something that I hope will help the thousands of users who can't stand the changes to the interface. Microsoft has disabled the ability to boot straight to the Desktop. So here's a workaround. 

Put a couple shortcuts to programs in the "Startup" folder. 

The path is C:\Users\username\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Startup

OR access the Run menu by pressing the Windows key and the R key at the same time then type %appdata% and browse to the folder above. Create shortcuts to Internet Explorer or a program in this folder. 

When Windows boots, you will still see the Start menu for about 20 seconds (depending on your computer speed) and once the items in the Startup folder are executed, the Start interface will close and go to the Desktop. 

This is best suited for someone who logs in and then goes to get coffee or reads physical mail, goes to the bathroom etc after typing their password. If you're going to be at the computer you could just click the Desktop button and go straight to it, but this is probably the best option for passing the Start menu right now.

"bypass start windows 8"
"bypass start menu windows 8"
"bypass Metro"

Dell PowerEdge 2850 RAID benchmark

In determining the RAID mode needed for a project I ran benchmarks using HDTune on a Dell PowerEdge 2850 with the PERC 4 controller on U320 SCSI 10K RPM disks.  Hopefully this can help someone out in making a decision on what mode to use. Unfortunately I didn't benchmark any other RAID modes besides 50 and 10.


I will not go into details on how RAID works in all it's incarnations here, there are tons of documents elsewhere on the web.


If you cannot see the picture below I have typed out the details in a table.  In short, RAID 10 offers superior performance in all areas except CPU usage, however with a 1% difference this is hardly a deciding factor. There is also a 146GB (1 disk) loss in a RAID10 configuration.

Ultra 320SCSI 10K RPM	RAID 50	RAID 10
Usable space	586 GB	440 GB
Transfer minimum	15.4 MB/sec	94.6 MB/sec
-Maximum	73.8 MB/sec	179.4 MB/sec
-Average	32.8 MB/sec	123.1 MB/sec
Access Time	8.7 ms	5.9 ms
Burst rate	256.3 MB/sec	291.2 MB/sec
CPU Usage	0.3%	1.3%

Windows share redirection using a DNS alias

In order to simplify this case, I've reduced the number of objects. In this example, there are 4 servers and 2 clients. The DNS/WINS system is Windows Server based and replicated between 1 other domain controller. The situation requires the retirement of one server and all of its shares migrated to another server. Specifically, the need for the DNS redirection is based on the requirement (the problem) that there are hundreds of files (Excel etc) that contain links to other documents using the full UNC path to the old server instead of a mapped drive.

Server A ("SA.network.com" / 10.1.1.2 Server 2003) - Running DNS & WINS
Server B ("SB.network.com" / 10.1.1.3 Server 2008) - Running DNS & WINS
Server C ("SC.." / 10.1.1.4 - Server 2003) - Old (to be retired) file server
Server D ("SD.." / 10.1.1.5 - Server 2008) - New file server
Client A ("CA.." / 10.1.1.6 - WinXP x86) - any user workstation with mapped drives
Client B ("CB.." / 10.1.1.7 - Vista x64) - any user workstation with mapped drives 

SC is sharing \\SC\share1. In preparation, SB has the same share created and permissions assigned. DNS has dynamic forward lookup zone entries for each server (as does WINS-which is linked to DNS. I will explain this later).

In our first attempt, the files were copied from SC to SD. As a side note Microsoft provides a utility for this called MS File Server Migration Tool which is very easy to use and carries over the share and file permissions. In our case this worked fine on the first server and on the second move we used BackupExec to restore the full and incremental backups which was twice as fast as using FSMT. The DNS entry for SC.name.com was changed to a static entry to the IP of SD - 10.1.1.5. Once DNS replicated between SA and SB then updated on the clients (either by it's normal time schedule or by manual flush) the new server would respond to ping on the old server's name, but would not work by UNC browsing by name (I believe using the IP worked but what good is that in this case?) Windows would report the error "a duplicate name exists on the network". Microsoft and other forums have tons of questions and responses about this error.

Our next step involved removing SC from the domain from the server by using System>Computer Name - NOT by deleting it from AD) This got us further, but not yet to a working state. Oddly, CB running Vista x64 was able to browse by share name. Why it worked there but not on CA I don't know. We then manually deleted SC's computer account from AD. It seemed that somehow the AD computer account was doing some sort of reverse lookup to the IP bypassing DNS but reflecting back on this case it was probably DNS to WINS. I'm no AD expert but this it what appeared to be the case.

At this point CA and SA still could not browse by name. The next process involved discovering that DNS was referencing WINS for entries it wasn't aware of. This is a setting somewhere in the DNS server properties. I can't recall how this fit into our situation but we had to manually delete (graveyard) the WINS entries. If you don't have WINS this doesn't apply.

After searching around this article (http://support.microsoft.com/?id=281308) came up which referencing share aliases and a registry entry for strict name checking. I don't have any technical details on what this function actually does but it ended up being the last step to resolve our problem. On SD, the registry change was made, restarted and browsing by name worked on all clients.

1. Copy files from old to new server
2. Disjoin old server from domain, then manually delete the computer account in AD.
3. Set static DNS entry using old server's name to new server's IP
4. Optionally ensure WINS isn't referencing any of the servers.
5. Make the registry change on the new server, restart.
6. Ensure DNS/WINS replication happens between any DCs or DNS server roles.
7.  Wait for DNS to update on clients, or manually flush.

Symantec Ghost 11 - General Protection Fault at eip=39edcf

In an attempt to reduce our VLAN count, we're attempting to merge one and all its devices into another. We've used Ghost countless times on the old VLAN but ran into an issue on the new VLAN.

In this example we have 2 VLANs:
1 - Router=192.1.1.1
2 - Router=192.2.1.1

The Ghostcast server (192.1.1.100) sits on VLAN 1 and the client is on VLAN 2. The client boots by CD into the PC-DOS Ghost application and is attempting to setup a unicast session to create a backup. When going to the Network>Unicast menu, Ghost immediately crashes when trying to get an IP address with:

General Protection Fault at eip=39edcf; flags=3016
(see the screenshot for the trailing error codes)

This error does not appear to be on Symantec's website and Google returned nothing (hopefully after this post it will).

This is because there was a mistake in the DHCP scope options. The "Router" address (option 003) was set to the VLAN 1 router instead of it's local VLAN 2. I don't know exactly where in the programming the crash occurs because the Ghostcast session isn't initiated yet, this happens during the DHCP lease process. Again, oddly even after changing the DHCP scope options to the correct address then running Ghost again from the command line without rebooting the program still crashes. Windows computers on the same network have been given this incorrect route and have been working fine, so I don't fully understand this. Once the DHCP scope option was corrected Ghost worked (after a reboot).

I figure this is a niche error because there's no reference to it anywhere and our network is likely unique.

Welcome

I've started this blog to post random tips and things that come to mind (usually technical subjects). I'm not a professor, but a friend of a friend saw me wearing a cardigan and said "Who's the professor?" One of those had to be there things.