We replaced gp with a combination of things. we ported dremel to our heavily customized hadoop (it is, even now, at least 10x more performant than the public h

We replaced gp with a combination of things. we ported dremel to our heavily customized hadoop (it is, even now, at least 10x more performant than the public hadoop code, drawing from experiences of running on 5k+ physical nodes, plus yahoo is where software engineering goes to die) we also ended up using spark while neither of the above is as convenient as greenplum could be, they also didn't have 2+day outages every other month while hundreds of tb of data had to be regenerated and reloaded for the reporting powered by greenplum. They also didn't cost millions of dollars a year that emc wanted. Overall, I think we were very happy ending gp.