Ryan, any reason, why you wouldn't simply have a RAID array with the drives : it will be faster, less wasteful, and easier to manage ...
Dali On Thu, Jul 30, 2009 at 6:54 PM, Amandeep Khurana <[email protected]> wrote: > Ryan, > > You can surely do this by having 4 virtual machines (I havent tried that > myself though). However, make sure you understand the principles behind > hadoop. The fundamental reason to go towards hadoop is to distribute the > load across multiple machines. Everything that hadoop provides is with that > basic idea in mind. Replication basically provides tolerance towards > failing > hard drives. With 2 hard drives, a replication factor of >2 wont add value. > Secondly, HDFS is slower than the filesystem you would run it on... So, > throughputs would be low. > > So, make sure you understand the design philosophy and then move forward on > it. > > -ak > > > > On Thu, Jul 30, 2009 at 6:15 PM, Ryan Smith <[email protected] > >wrote: > > > Sorry if this has been asked already, I didnt find anything on google. > > > > I want to run a dfs cluster on one machine. 8 core processor, 8 GB RAM, > 3 > > x > > 2TB drives. I want triple replication on the dfs. > > If i make 4 IP aliases on the machine, (10.1.2.30,10.1.2.31, 10.1.2.32, > > 10.1.2.33) can i run 1 master (NN & SNN) and 3 slaves (3 DN)? > > Or what is the best approach? I want to make a private hadoop storage > for > > myself, but i dont want to buy a bunch of servers. > > Thanks, any advice is appreciated. > > -Ryan > > > -- Dali Kilani =========== Phone : (650) 492-5921 (Google Voice) E-Fax : (775) 552-2982
