Data Sharding or Master and Slaves

I am coming in on the tail end of a db optimization plan provided by our software/database vendor. 

The problem: Poor performance on writes and too much data in certain tables which results in poor results, long query times on certain reads. 

The solution: Data Sharding or Master for writes, Slaves for Read and Master Slave replication. 

I am new to both ideas and have been doing research on them this week. Both feel like good ideas but I am not a DBA type; I am a software guy who knows enough to get by on the db side. 

For those that have looked at both as a solution to poor performance and too much data (tables are growing very fast; good news as the business is growing very fast) can you share your experience? Which did you go with and why? 

Also, I have fond what I think are some good links on the subjects though I am interested in any information folks can share with me on this. 

Thanks!