The .NET Framework offers two methods for rounding: Decimal.Round() and Math.Round(). However, both methods implement a type of rounding known as "banker's rounding," which rounds .5 to the nearest even number. This is a different implementation from the "standard" arithmetic rounding most people are familiar with, where .5 rounds "up." Even the definition of "up" can vary, depending on whether the rounding is symmetric or asymmetric. Symmetric rounding rounds in the same direction relative to 0 (either towards or away). Asymmetric rounding always rounds in the same absolute direction (towards positive infinity or negative infinity).

Banker's rounding has its benefits, but it can cause incompatibility with other systems that implement different rounding schemes. For example, Math.Round( 8.5m ) will round to 8 in .NET, but round( 8.5, 2 ) will round to 9 in T-SQL. Inside is source code that implements arithmetic symmetric rounding in .NET, the type of rounding that T-SQL uses.

See How To Implement Custom Rounding Procedures for additional information

/// <summary>
/// Implements alternate rounding methods, in contrast to Decimal.Round()
/// and Math.Round(), which implement banker's rounding (5 rounds to even).
/// See http://support.microsoft.com/?kbid=196652 for other methods.
/// </summary>
public static class Round
{
  #region Methods

  #region ArithSym
  /// <summary>
  /// Rounds using arithmetic (5 rounds up) symmetrical (up is away from zero) rounding
  /// </summary>
  /// <param name="d">A Decimal number to be rounded.</param>
  /// <param name="decimals">The number of significant fractional digits (precision) in the return value.</param>
  /// <returns>The number nearest d with precision equal to decimals. If d is halfway between two numbers, then the nearest whole number away from zero is returned.</returns>
  public static decimal ArithSym( decimal d, int decimals )
  {
    decimal factor = Convert.ToDecimal( Math.Pow( 10, decimals ) );
    int sign = Math.Sign( d );
    return Decimal.Truncate( d * factor + 0.5m * sign ) / factor;
  }
  #endregion ArithSym

  #endregion Methods
}