Author Archives: trainridetothecity

MPEG DASH the new kid in HTTP delivery

Posted on by

image

With my experience working with hundreds of handset vendors, Apple has been most challenging. This is not surprised since they took smart phone market in to a new era leaving competition light years behind. Market leaders such as Nokia had to leave Symbian behind and join forces with Microsoft. And handset manufacturers such HTC, LG and Samsung saw Google Android coming to their rescue.

Until Apple came with HTTP live streaming protocol with iOS 3.0, RTSP was the only supported live streaming for feature phone and early smart phones. It was built for real time delivery with all the tools needed to optimize fast delivery to end devices. RTP payload was on UDP and over the top session management was on TCP with HTTP like RTSP protocol. It was a nightmare for over the top  providers who did not control the mobile network with NATing an Firewalls. But traditional telcos built their network around these challenges.

But Apple made it a level playing field with Apple HTTP live streaming protocol. Protocol is simple you chunk the live stream into small files, create a running playlist pointing links to the segments. Put a CDN infront to cache these playlists and segments. Everything is on network friendly HTTP.

It was great until Microsoft built Smooth and Adobe built HDS. The technology got fragmented with device specific deliveries. This became a nightmare for us. Supporting three different HTTP deliveries with layers of DRM technologies.

Now the industry is changing with MPEG releasing a standard which has a acronym of MPEG DASH.

It is an interesting time to be in the media delivery world, and very interested to see how big guns like Microsoft, Apple and Google play their cards.

Pixels per inch

Posted on by

At work, I have been asked to demo PPI (Pixels Per Inch) value on different video profiles and its effect on the iPad.

Well what is the PPI and how do you calculate that? It was quite simple at the end


W=Width of the video
H=Height of the video
S=diagonal size of the screen
PPI = SQRT(W^2+H^2)/S

Why does this matter? When video becomes available to handhelds such as tablets, broadcasters want to ensure that these new devices are not becoming a substitute to traditional broadcasting experience on large panels at your home. Thus PPI limitations are enforced to ensure video’s available on tablets are not mirrored to your TV screens at home.

Finally it has started

Posted on by
image

After two days of cleaning up. two trailer loads of rubbish and concrete

Finally after 7 months waiting time the site work has started on our new house.

Metricon has started work on 24th October with earth works as promised. To be honest this is the only thing they have done on time. Its an another story. We have been to the site almost everyday now.

24th October: Site start with scrape and leveling
25th October: Some new pegs, seems something has happened. but can’t figure out what!

image

27th October: We have electricity! Looks like some kind of piers has been installed and poured some concrete.

Our site supervisor Josh hasn’t called us yet, but it’s good to see something happening on the site.

Display Twitter Feed on Polycom

Posted on by
Twitter feed on the Polycom

Twitter feed on the Polycom

Well I got this new IP Phone replacing our old PSTN phones. But the LCD screen on the phone looked bit boring.

There were few ideas how I can use this realestate to do something interseting.

  1. Display the train timetable (Metlink doesn’t give you a clean API to scrape the time tables, and I don’t have time to write a script scrape the data from its bulky website. )
  2. Weather (Kind of boring to watch it all the time)
  3. What’s people talking about your company on  Twitter, and thats what I did…..

My Ugly Bash Script

#!/bin/bash
text=`export http_proxy=http://xx.xx.xx.xx && /usr/local/extractSearch.pl`
convert -size 160x75 -monochrome -depth 1 -type Bilevel -gravity center caption:"Twitter Feed\n$text" /var/www/polycom/image.bmp

The Perl Script
#!/usr/bin/perl
# use module
use XML::Simple;
use LWP::UserAgent;
use Data::Dumper;
# create object
$xml = new XML::Simple;
my $url = "http://search.twitter.com/search.atom?q=<Your Query>&rpp=1&page=1&locale=en";
my $ua = LWP::UserAgent->new();
$ua->env_proxy;
$ua->agent('Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1)');
$ua->timeout(30);
$ua->default_header('pragma' => "no-cache", 'max-age' => '0');
my $res = $ua->get($url);
$data = $xml->XMLin($res->content);
print "$data->{entry}->{title}"

Am I hacked?

Posted on by

I found a very strange thing when I saw DNS queries for random servers on Wireshark. I am running Ubuntu 10.10 and I was pretty sure that no one is running malicious scripts on my Laptop. I was really worried when I saw DNS queries to http://www.ssa.gov and I had no idea what it was at the time. And when I googled to find out what it was the DNS queries increased.

So I had to get the bottom of this. “lsof” is a very useful program to see which process is responsible for the which network transaction. When I ran the following I found the culprit.

poo@SDT-TravelMate-8572G:~$ lsof -i udp
COMMAND    PID    USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
firefox-b 1855 poo   42u  IPv4  27548      0t0  UDP SDT-TravelMate-8572G:41473->dns1.xxx.com.au:domain
firefox-b 1855 poo   62u  IPv4  27547      0t0  UDP SDT-TravelMate-8572G:35432->dns1.xxx.com.au:domain
firefox-b 1855 poo   63u  IPv4  27556      0t0  UDP SDT-TravelMate-8572G:34421->dns1.xxx.com.au:domain

So the culprit is Firefox doing DNS pre-fetching. Interesting article on that

 

Linux Bare Metal Restore

Posted on by

At work I have been given the task of coming up with a procedure of bare metal restore of redhat servers. There were two requirments.

1) Hot backing up
2) Remote restore

After some researching this is the procedure I came up with. Please bear in mind this procedure has only been tested in model enviroment and do your own homework before using this as your DR plan.

A) Backup

Backup the mbr and the bootloader. First 512 bytes of the primary HD contains the partion table and the bootloader. Here is the command for that.

dd if=/dev/sda of=/backup/mbr.img bs=512 count=1

Ensure that you have the backup directory and correct HD.

Now we need to backup the filesystem. Linux is dynamically creating the following directories. So no need to take the backup of them.

1) /sys
2) /proc
3) /var/run
4) /lost+found
5) /tmp

You need to exclude the backup directories. You can use tar or rsync to take the backup, however if you use rsync ensure you preserve symlinks and file permissions. Here is the command to create the tar archive.

tar cvfpz /backup/root.tar.gz –exclude=sys –exclude=proc –exclude=var/run –exclude=tmp –exclude=lost+found –exclude=backup /

Take a backup of the /etc/fstab and partion table info from the following command.

fdisk -l > /backup/fdisk.txt

B) Restore

Okay before restore lets crash the HD. Please beware this command shall ruin your HD, do not run this in the production enviroment.

dd if=/dev/random of=/dev/sda bs=512 count=100

Okay let’s start restoring.

You need a lInux live cd. Well I prefer a knoppix since it has all the useful tools.

Boot the server with the knoppix cd and restore the partion table. Since I am soundf this remotely I will be using iLo virtual media to mount the knoppix live cd.

dd if=/backup/mbr.img of=/dev/sda bs=512 count=1

Linux will not identify the HD until you reboot, however you can do the following as well. Run fdisk and write the partion table by pressing ‘w’.

Okay now let’s format the partions. This is where the copy of fstab helps. You can run the following command for each partion.

mkfs.ext3 /dev/sda2 -L /
mkfs.ext3 /dev/sda1 -L /boot

Now your partions are ready for the data. Let’s mount the devices. First create the mount point by running the following command. I am killing two birds with one command. I am creating rootfs and boot directories.

mkdir -p /rootfs/boot

Okay change directories, into cd /rootfs

Let’s extract the archive

tar xvzp /backup/root.tar.gz

Okay now we need to create the following directories.

mkdir /sys; mkdir /tmp; mkdir /lost+found; mkdir /proc; mkdir /var/run

Okay now you need to chroot into the /rootfs

mount -t proc /proc /rootfs/proc
mount -o loop /dev /rootfs/dev

chroot /rootfs -c /bin/bash

Now you have mounted the filesystem. Run the following command to reinstall grub

grub-install /dev/sda

Okay fingers cross you have recovered your crashed server. Reboot the box you have resurrected your Linux box.

Things you need to know. This won’t work if you have a LVM. If your fstab has uid rather than block devices, you need to change it to block devices.

If you have grub2 grub reinstallation might change a bit.
I will cover that in another post.

Okay something you might notice that I didn’t cover the backup directory. You have several options here. You can use a nfs or samba mount for that. Please keep in mind if you are using rsync as your backup method you cannot have the symlinks in a samba mount. So you need to use tar if you are using the samba option.

If you have a better way or see a mistake please comment.