In Part 1 and Part 2, we used the MapR Converged Data Platform to plan a real-life Greco Player Tracker (GPT), the AI security system protecting the Bank Hotel Casino in the 2007 film Oceans’ Thirteen.
The hero of Oceans’ Thirteen isn’t the casino or the GPT, however. The heroes are Danny Ocean and Rusty Ryan, who need to beat the AI in order to rob the casino. They already know that the GPT:
- can’t be beat, hacked, or simply unplugged
- is kept in a temperature controlled, shock-resistant room
- has a reboot time of three and a half minutes
In the movie, they decide to take down the system with a magnetron to force shutdown and reboot, giving them a three and a half minute window to rob the casino. But would this be possible on our GPT built with MapR?
A 200-node cluster with an exabyte of data stored across 4800 disks would take quite a while to restart. It may take some time to physically get the disks spinning again. After a hard shutdown caused by a power outage or a magnetron, Linux and other Unix-like systems may take two or three minutes to reboot, as the OS checks for errors. Warden and Zookeeper may take up to a minute to start up as well. Finally, other services like MapR Streams or Spark Streaming may take some time as well.
Conservatively, we can expect a reboot time of five minutes. This means Danny and Rusty would have 90 seconds more than they had in the movie to rob the Bank Hotel Casino.
The longer reboot time does not mean the GPT with MapR is easier to break, however. Danny and Rusty spent millions on drills and the magnetron just to break into one location. With MapR, replication to other data centers means that our program can keep running, even if one data warehouse is taken down by a magnetron.
Since we mirrored our clusters in two off-site locations, the thieves would need to coordinate three simultaneous attacks to take the whole system down. This means that they would also need to triple the millions spent on the drills and magnetron, thus greatly reducing Danny and Rusty’s bank robbing ROI. Furthermore, they would need two additional teams of people to take down the off-site data centers. Whatever money they can steal in five minutes will, in this scenario, have to be split among more people: a real ROI killer.
Although the three and a half minute reboot time from Ocean’s Thirteen is not realistic, MapR’s replication and mirroring make breaking the system much more difficult. The security provided by MapR would increase the cost and decrease the ROI for any thieves wanting to break into our system.