Vdbench50403rc9: updates to Vdbench50402

Question

Attached: vdbench50403rc9.zip
Zip file includes PDF.
I expect, or better yet, hope, that this will be the next GA version Vdbench50403.
I just want to make sure that I have not broken anything, so I'll wait to see if any new or old problems show up before I declare this GA. So please give this a try. Keep in mind, this is still not considered 'done'.
Changes since 50402
Vdbench now requires java 1.6 or higher.
Data Validation -vt or 'validate=(yes,time)' option did not store last time read, only last time written.
The aforementioned -vt option will now report 'last time block was valid' in milliseconds instead of just in seconds.
If a Data Validation pre-read failed and the 'read immediately' -vr or 'validate=(yes,read)' option was used the corrupted block was reported twice.
False corruptions when SDs or FSDs were allowed to move between slaves.
'slow socket message' display has moved from the console to logfile.html.
When running with a large amount of SDs and slaves the amount of detailed performance statistics transferred each interval from the slaves to the master could become overwhelming resulting in 'slow socket' messages. Since details response time histograms were only reported after normal termination sending this huge amount of information to the master was a waste of time. Histogram data therefore will only be sent over at normal termination. This fix should have a positive impact on those test runs where the detail interval statistics are not reported in time.
New option: messagescan=nnnn, where the scan for messages in /var/adm/messages is terminated after 'nnnn' messages. Default is to stop scanning and therefore reporting after 1000 lines.
To continue making it clear that Vdbench is owned an distributed by Oracle, a copyright statement is now displayed each time Vdbench is executed: one on the console and one on summary.html and logfile.html. This is an extra reminder that companies that distribute Vdbench parameter files are indeed distributing parameter files only but that they should not imply that they also own or distribute Vdbench.
Java master/slave socket connections using port 5570 are now released as soon as the master/slave connection has been established, freeing up this port for other Vdbench runs. This means that earlier default limit of 10 concurrent Vdbench executions now is virtually unlimited.
New file selection option: fileselect=(poisson,3) allowing for a more skewed control of which files will be used. That comes in handy when you have millions of different files.
New 'format=' option: format=(no,complete) where you tell Vdbench to 'trust you' and that the format indeed has been done but that there is no need to do a huge amount of directory searches.
Fixing a serious bug that in Vdbench50401 under certain circumstances would NOT honor requested workload skew. The fact that the skew failed was still reported however.
A new 'skew.html' report that now gives a lot of detailed information about the amount of work done for an SD, WD, Host, or Slave. And this of course because of the previously mentioned bug.
A new general parameter: abort_failed_skew=nnn where Vdbench will abort if the workload skew misses by nnn%.
'Shared threads' when using concatenated SDs: though the objective has always been for the rd=xxx,….,threads=nnn parameter for concatenated SDs to be a TOTAL amount of SHARED threads, the code erroneously treated the value as 'number of threads per workload'. This has been corrected.
Workload skew control only works when all workloads 'compete' with each other on every slave/JVM. There were scenarios due to specified slave and thread counts that not every workload ran on every slave. Vdbench now will dynamically round thread counts upwards so that this no longer can happen.
GcTracker: Information will be displayed on each slave's stdout.html report about the amount of java Garbage Collection that has been done. This usually is not a problem until you start working with millions of directories and/or files.
New file system 'operation=access' functionality.
File system output now will ONLY display the columns for getattr/setattr/access if those operations have been specifically requested.
Status.html: an 'official' API to be used who want to monitor the current status of Vdbench, typically some 'wrapper' around Vdbench.
OSX: OSX does not use the 'normal' flags that other types of Unix can pass during a file open() call. They use fcntl() instead. Vdbench has been fixed to now use fcntl() ONLY for OSX.
Vdbench now also has been tested on Raspberry Pi (This was just a fun little project).
I updated the documentation to the best of my ability to include everything that has changed.m

Answer 1

FIXED!!!
In addition to the other things I previously mentioned:
Turned OFF Wifi (Settings>Wi-Fi>Off)
Turned OFF Dynamic Wallpaper (Settings>Wallpapers & Brightness>Choose a wallpaper from the Stills collection)
Turned ON Reduce Motion (Settings>General>Accessibility>Reduce Motion>set to ON)
Turned OFF Background App Refresh (Settings>General>Background App Refresh>Off)
Turned OFF Location Services (Settings>General>Privacy>Location Services>Off)
Manually kill all running apps frequently (Double tap HOME button, swipe running apps UP to close)
Set screen brightness very low (Settings>Wallpapers & Brightness>set slider about 50-60%, make sure Auto-Brightness is ON)
I found that disabling Mail Fetch stopped the ridiculous battery drain.
Settings>Mail, Contacts, Calendars>Fetch New Data>Push ->set to disabled
Now my iPhone 5 has about the same battery life as I experienced before the upgrade. I can still go into Mail and it will update just fine like before.
Hopefully this helps. I'm hoping this is just temporary and Apple fixes this with 7.0.1.