On Mon, Aug 25, 2014 at 1:44 PM, Srinath Perera <srin...@wso2.com> wrote:
> Hi All, > > BAM capacity planning it bit involved than other products. I was doing one > such case with Mifan and following are steps we followed. Please comment. > > 1. Need to know roughly how much data you need to handle. One way to get > this is knowing the event rate. Lets say 100 TPS. That means if each record > is 2K, then it is about 18GB per day, about 550GB per month, and about > 6.5TB per year. > > 2. First Questions is how long we keep data. Generally, this should be > like 1 month. You need enough disk to hold the data. If you have 3 replicas > you need 3X that, and Cassandra also add an overhead ( > http://www.datastax.com/documentation/cassandra/1.2/cassandra/architecture/architecturePlanningUserData_t.html > ) > > 3. Then think how much data you will be processing with Hive. If you > process once a day, each processing will pull like 18GB. Need to make sure > there is enough computing power. Hadoop is mostly decided by number of > cores, and Cassandra we follow data stacks recommendations. I think core > can do about 20MB second (only a guess). See > http://blog.cloudera.com/blog/2013/08/how-to-select-the-right-hardware-for-your-new-hadoop-cluster/ > . > > --Srinath > > > > > > > -- > ============================ > Srinath Perera, Ph.D. > http://people.apache.org/~hemapani/ > http://srinathsview.blogspot.com/ > -- ============================ Srinath Perera, Ph.D. http://people.apache.org/~hemapani/ http://srinathsview.blogspot.com/
_______________________________________________ Architecture mailing list Architecture@wso2.org https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture