Data integration solutions that perform well, especially on "big data," evolve from good design. These solutions exhibit one or more of the following characteristics: high level of parallelism, maximized bulk operations, reduced network data overhead, and an optimized/ nonduplicated transformation layer. While these considerations look straightforward, unfortunately, there is no silver bullet. A design solution is context-specific and is often a combination of the above characteristics.