Its been a busy day for both Twitter & Rush Hour powered Tweetrush. Thats right, the first million barrier has been busted. Check out the Tweetrush stats for 27th August 2008.
While congrats go to Twitter for coping with the extra traffic while the DNC was on, congrats also go to Gnip, our feed provider and the Rush Hour team for desiging such a stable system considering the traffic and data we get and analyse. Its no mean thing to do, and I see the DB’s that are involved so its certainly a lot of fun.
I’m off line for a few days while I attend a family event, however I promise to respond to all RH mails when I come back. In the meantime here is a another great article on a use for Tweetrush.
Well, what an exciting few days it has been here. TweetRush.com, the first Rush Hour proof of concept site is up and running along and so far seems to be holding fast. I’m taking a break tonight and actually sleeping for a change but rest assured all mails and tweets will be answered.
What is TweetRush about? Well its a demo of a new product we’re building that aims to take the pain out of stats for web app builders and is based on actions and events. But more on that will be released as we get closer to things getting out the door.
We put TweetRush up to get an idea of how well the engine would perform against a high volume site. Sure there are few issues mostly around the fact that we don’t get a direct Twitter feed, but over all its not bad. We hope once all is calm, to establish a good contact with Twitter and maybe a XMPP feed and see what we can both come up with. Do remember, this is still a work in progress
Now Rush Hour and the Tweet Rush spin off are not just my babies, but also James, Walter, Grzegorz and Slawomir babies too. We have all equally put a massive effort in to get this far. Its been great to be part of a very talented team. There is also Adam, but he was busy doing marriage or something crazy so we excused him for a while, no escaping us now Adam now that you know we mean business
Without all of these great people involved we would never have gotten this far at all.
I am also delighted to say we featured on TechCrunch, thank you for the exposure Michael Arrington! Thats another first I think for all of us on the team. Well since then, we’ve been inundated with greetings and well wishes as well as many questions too. Site accesses have gone through the roof (and a big thank you Donncha for the advice on what to expect DB wise). All feedback has been great to get, allowing us to enjoy the moment, but also to look deeper into our application and the Tweet Rush implementation to see what else we can do to improve its accuracy as well as seeing what peoples expectations are.
Finally there are four other people I must say thank you to as without them, nothing would have happened either. Patrick Buttimer of Eirteic Consulting for being just absolutley great. Justyna for being there and not walking out after I deserted her for weeks to work on systems and code. She’s a babe that rocks! Damien, of Mulley Communications, the PR was excellent, and support through the highs and lows, nerves and all has been second to none. Finally the growing legend that Pat Phelan of Twitterfone/Max Roam/ … (endless list of great businesses) is. His advice via tweet tennis (Is that a new term Damien?) and help in reaching people has been great. A real genuine gentleman.
Okay so sleep beckons now, I’m really hoping nobody diggs us now, being TC’d is enough for the one day
More to come on Rush Hour later after I sleep.
P.S. For the PHP peeps, of course Tweetrush was written in PHP using the Zend Framework
Seriously it rocks.
These stats were captured via the impending Rush Hour engine that myself and a few others have been developing. More info on where they come from later in the week, or eary next week. For the moment, lets just say big DB’s rock, even bigger machines rock harder and see if we can guess where they come from.
For every action there can be an equall reaction!
So I’m doing something that relies on MySQL a lot. Last night I had to test a rebuild process that we would use if there was a failure in our systems. Worst case secnairo would be we would have lost about a weeks data and need to reinsert the whole lot.
So I wrote a quick and nasty PHP script with multiple nested foreach() loops, yeah I know what your thinking, but as I said, it was a quick and nasty script. Basically all this script done was parse a directory full off XML files and based upon the content, perform multiple SELECT, INSERT or UPDATE based on the data per element in the given XML file.
So, on my little iMac (24″, 4GB RAM, 1TB Drive) the following are the stats from the completition of the script.
Total Processing Time: 4975.59566307 seconds.
Total DB Inserts: 1,961,000
Total Selects: 12,035,743
Total Updates: 10,465,071
This all translates to;
Inserts Per Second: 394.12
Selects Per Second: 2,418.95
Updates Per Second: 2,103.28
Keep in mind, I was watching a movie at the same time and that there has been no optimisation carried out on the server at all (Although the DB is fairly well normalised). Table format was MyISAM.
Now imagine what MySQL could do on an optimised Linux server with custom optimisations to make it faster. I am quite impresed given the workstation I was working on and the complexity of the queries we are doing. Pat on head for MySQL. Well done.
Comments