Corrosion coupons may not be useful for predicting data center equipment failure rates

Information Technology (IT) equipment manufacturers have claimed that there are increased failures of IT equipment attributable to gaseous contamination, although there is little quantitative evidence of this. This trend reportedly has occurred due to a switch from leadto silverbased materials due to the Restriction of Hazardous Substances Directive (RoHS) mandate of 2006. The industry is considering putting more importance on using silver corrosion measurement coupons as a potential failure indicator due to the unconfirmed belief that silver coupon measurements may better predict the failure of electronic equipment. We performed an exploratory study of 19 data centers in the United States and 2 in India. Coupons containing one copper and one silver strip were deployed for 30 days in each data center. Coupons were placed in the supplyair path starting where the air entered the building, inside the plenums and ducts feeding the data center rooms, and inside the data center rooms. The goal of the study was to investigate the following questions: (1) What are the approximate statistical distributions of copper and silver corrosion in data centers across the U.S.? (2) Are corrosion coupon measurements repeatable? (3) What is the relation between copper and silver measurements? (4) Are corrosion rates higher for outsideair cooled data centers compared to "closed" data centers? (5) Are corrosion measurements related to IT equipment failure rates? The results of the study indicate: (1) Copper and silver coupon corrosion rates were generally low in the U.S. compared to rates that are thought to be problematic. (2) Measurements within the same data center frequently differed by a factor of 2 or more. (3) Silver corrosion rates were poorly correlated with copper corrosion rates. (4) Copper corrosion rates are not higher for outsideaircooled data centers than for "closed" data centers. Silver corrosion rates are not higher in most aircooled data centers, but may be higher in some. (5) Data centers with relatively high silver corrosion rates reported no unusual equipment failure rates, although detailed analysis of this question is not possible because of the low number of data centers with high silver rates encountered.