Fleming, I'd add integration with Hadoop & MapReduce support for data mining tasks as one of the major pros.
On the other side - the size and complexity of the code base/API/Configuration is something to consider. Look at the size of the source code in KLOC, number of hadoop/hbase/zookeeper configuration parameters etc. Hadoop/Hbase project is a bit of a heavy-weight in this regard. See if you can make sense of these parameters before deploying into prod. compare to similar or more lightweight systems which might do the job, you will have to support it after all. i'd also recommend running this benchmark on your hardware before making any decisions: http://wiki.github.com/brianfrankcooper/YCSB/ Cheers Alex http://www.columbia.edu/~ak2834/ <http://www.columbia.edu/~ak2834/> On Mon, Jun 21, 2010 at 11:07 PM, Todd Lipcon <[email protected]> wrote: > Hi Fleming, > > Lots has been written about this if you look through the archives. Just a > few notes below: > > 2010/6/21 <[email protected]> > > > Hi there, > > > > I would like to get a few words about HBase's pros and cons > > that may probably help my boss to make decision of adopting > > HBase as production. > > > > Pros : High volumn data random access > > Scale-out with commodity machine > > Fault-tolerance > > Free license > > > > Cons : No security control > > > > Andrew Purtell and his team at Trend Micro are working on this in the next > quarter or two. Please refer to: > https://issues.apache.org/jira/browse/HBASE-1697 > > > > Data loss risk > > > > This is essentially fixed in our next major release, assuming you are > running the right build of HDFS. More coming to the user list next week on > this subject, but with proper sync() support in HDFS, data loss should > never > happen unless there are bugs. Bugs that do cause data loss will be treated > with highest priority. > > > > Redesign data schema > > Lacking of aggregate function(Max, Min, Avg...) > > Multiple client concurrent read/write performance > > > > Not sure what you mean about this - there have been some performance bugs > in > the past with contention, but we've improved and will continue to improve > on > performance. > > > > No commercial support now > > > > > Cloudera is beginning to offer commercial support for HBase with CDH3. Let > me know off-list if I can put you in touch with our sales people (I don't > want to make the community list a sales forum!) > > > > > Any suggestion or correctness would be appreciated! > > > > > > Fleming Chiu(邱宏明) > > Cloudera Certification for Hadoop Map/Red Developer > > TEL: 707-2260 > > Email: [email protected] > > Be Veg! Go Green! Save the planet! > > > > > > > > > --------------------------------------------------------------------------- > > TSMC PROPERTY > > This email communication (and any attachments) is proprietary > information > > for the sole use of its > > intended recipient. Any unauthorized review, use or distribution by > anyone > > other than the intended > > recipient is strictly prohibited. If you are not the intended > recipient, > > please notify the sender by > > replying to this email, and then delete this email and any copies of it > > immediately. Thank you. > > > > > --------------------------------------------------------------------------- > > > > > > > > > > > -- > Todd Lipcon > Software Engineer, Cloudera >
