Is NoSQL really necessary or do we just need a better SQL?

Thomson’s  &  Abadi’s  proposal   As  a  solu?on  to  these  problems  Thomson  &  Abadi  propose  the   following:     1.  There  exists  a  scheduler  which  defines  an  a-­‐priori  sequence   for  the  transac?ons.     2.  Each  transac?on  has  to  acquire  all  locks  at  the  beginning.   3.  Aborts  are  not  allowed   4.  The  distributed  sub-­‐transac?on  behave  completely  iden?cal   in  all  nodes   These  requirements  indeed  increase  determinism  of  SQL   databases,  but  at  the  cost  of  adding  a  single  bo6leneck  that   nega?vely  impacts  scalability.  Statement  (5)  now  becomes   doubgul  at  least.