Now that we are rushing into the holiday season groaning slightly from the excesses of Thanksgiving but still licking our lips in anticipation of yet more turkey and stuffing, my thoughts are drawn to considering Santa’s IT setup. I mean, just think of it: How could Santa do his job without some serious IT support?
I’m sure that hundreds of years ago efficiency and productivity weren’t much of a problem for him, but now the world’s population has grown enormously and toys are more complex than they were. So it is no surprise that Santa now needs support for HTML (Holiday Toy Management Language), XML (Xmas Making of Lists), CSS (Christmas Stocking Stocking), and SOAP (usually Dove but anything that smells good and is wrapped in pretty paper will do).
So, what is the scale of Santa’s enterprise IT needs? Well, according to the Population Reference Bureau, about 30% of the world population (currently about 6.396 billion) is younger than 15. Assuming that 50% of the world’s children believe in Santa Claus, roughly 15% of humanity are eligible for gifts – a total of about 960 million children.
If each child’s database record contains his or her full name, say, 128 characters, address of another 128 characters, naughty/nice status flag of 1 byte, a list of presents he or she has previously received, say, 1,024 bytes for the list of UPC numbers (that’s Unified Polar Code), and a few control fields, then Santa has a database of about 2T bytes. Given the database is mission critical he probably has a hot standby; we’ll call that 4T bytes.
Add to that the data required for production of two toys per child (that’s just less than 2 billion toys) and supply chain, warehouse and facilities management, plus historical data. Then there’s elf resources management; if each elf can produce a toy per day and they work for 364 days per year – they’re non-union – there must be approximately 5.5 million elves on the payroll. So there has to be 8T to 10T bytes of very fast SQL database-driven storage.
Then there’s Santa’s logistics operation, which must have a supercomputer or two up its sleeve because, with 960 million stops to make, Santa needs some serious route optimization; this is the Traveling Salesman Problem (TSP) from hell.
Georgia Tech holds the record for the largest TSP solution, calculating the shortest route for visiting all 24,978 cities in Sweden. It took the equivalent of roughly 84.8 CPU years on a single Intel Xeon 2.8-GHz processor (the school actually used background processing on a cluster of 96 dual-processor Intel Xeon 2.8-GHz workstations between March 2003 and January 2004).
Assuming Santa needs an expeditious way to hit his 960 million stops but will accept a partially optimized solution, he might allow the problem to be divided into subsets the size of Georgia Tech’s solution and optimize each one individually. If the NPIT (North Pole IT) breaks the problem into subsets of 24,978 locations each, there will be almost 38,843 routes to solve. Further, if we assume that foreground processing would take one-fifth the time that the Georgia Tech problem took, NPIT would need about 8,000 clusters similar to Georgia Tech’s.
If each machine occupied a 1U rack space and assuming NPIT uses 50U racks, there would be 16,015 racks or roughly 128,000 square feet including access and power (around 120 megawatts of power would be required). And that’s just for route optimization!
Yep, Santa’s IT guys have their work cut out for them. Their job is not only mission-critical but they have to deal with enormous scale and a 12-month business cycle. Assuming a 1-to-200 IT-to-staff ratio, there’s at least 30,000 people in IT alone.
So, when the number of users you have to deal with and the business problems you have to solve are driving you mad, just think of Santa’s IT department and enjoy your turkey.
Ho-ho-ho to backspin@gibbs.com.
