I wonder why they went with something like this rather than a cluster like Percona's XtraDB Cluster that uses Galera replication? You could hang some read-only replicas off of the masters and achieve a similar system?
IMHO, it depends, I assume they mention about NDB Cluster not InnoDB Cluster. If their query using very less normalization, me myself also would choose MySQL Cluster instead.. their concept of auto partition + auto sharding did improve compare to other RDBMS. Plus I believe they should really understood the benefit of Index Condition Pushdown concept which really help a lot on complex query. Just remark, im not from MySQL but already tested several POC comparison on both DB, still MySQL Cluster did stand out a bit yet like i said depends on your transaction design back
