We’ve just started on Road Tour 2010, at the first stop in Montana to take care of business before moving on to Albuquerque for Convergence 2010, a conference of the Handweavers Guild of America. Summer also being Road Maintenance in the American West, I was struck by the similarities between computer networks and the highway system.
Most people assume the Interstate Highway system is a High Performance system, since it has multiple travel lanes in each direction, permitting parallel travel at different speeds without “blocking.” Parallel computer systems, whether Symmetric Multiprocessors (like multi-lane highways) or clusters (like the street grids in cities or concentric beltway systems) provide multiple paths for program execution, so there is no waiting in queues or waiting for an opening in the opposite lane to pass.
But, during Road Maintenance season, it is painfully obvious that the Interstate highway system is High Availability instead. High Availability systems do some load sharing, so appear to be High Performance, but the real purpose is to carry on the job if one of the nodes fails. A lot of our four-lane interstate highways end up being alternating sections of two-lane highways, as road crews perform necessary repairs to and replacements of paving and bridges.
Interstate 90, which traverses the northern US from Seattle to Boston, suffers greatly from the fierce winters in the Rocky Mountains and Great Plains. In the long winters, heaving asphalt, spalling concrete, and cracks from the inexorable slip of the roadways downslope result in poor traction indeed as your tires hover a fraction of an inch above the rough surface. After a few harrowing S-turns through steep mountain canyons, the narrow two-lane corridors through the construction zones are most welcome.
So, too, do computer systems need maintenance. The roads in Montana require it so frequently that the department of transportation has constructed the equivalent of hot-swap disks and power units, great paved permanent X’s between the numerous overpasses and bridges to facilitate switching traffic from four to two lanes and back. A high-availability computer system includes not only fail-over standby and load-sharing systems, but redundant components that can be replaced without ever turning off the system. The system administrator’s job is to monitor the internal failure sensors and replace failing items before the fall-back components also fail. Information must flow, even as the daily commerce of the Interstate highways must keep flowing.
In the Rocky Mountain West, there are often no parallel paths for detours: keeping the roads open in bad weather and during maintenance is as essential to commerce as keeping the company web server online 24x7x365. Those fast four-lane highways are really just redundant two-lane roads: take time and enjoy the scenery. Oh, and do check on the server status back home when you get to the next WiFi hotspot.