Can anyone comment on the suitability of databases of any kind as the data source for running a "high frequency trading" backtest? The fact that the market data ticks are in a database doesn't seem that helpful, given that the order book state needs to be built up by processing all previous ticks since market open anyway.