Schema advice: (Single row or multiple row!?) How do I store millions of columns when I need to read a set of around 500 columns at a single read query using column names ?

[Ertio Lew]

I want to store hundred of millions of columns(containing id1 to id2
mappings) in the DB  at any single time, retrieve a set of about 200-500
columns based on the column names(id1) if they are in single row or using
rowkeys if each column is stored in a unique row.


If I put them in a single row:-

- disadvantage is that the no of columns is quite big, that would lead to
uneven load distribution,etc.
- plus factor is that I can easily read all columns I want to fetch using
col names doing a single row read


But if I store them each in a single row:-

- I will have to read hundreds of rows(300-500 or in rare cases up
to 1000) at a single time, this may lead to bad read performance(!?).
- A bit less space efficient


What schema should I go with ?

Cassandra Hector failover + load balancing not as expected with version 1.0-5

[Roshan]

Hi

Currently I am using Cassandra 1.0.6 with Hector 1.0-5 in production, and
now in the process of upgrading current Cassandra cluster to 1.1.2 version.

As the first upgrade process, I select one node from local development
Cassandra cluster (3 node 1.0.6 and RF=3) and execute the nodetool drain to
empty the memtables and commit logs. During the drain, that node receive
reads+writes from external java client (which uses hector 1.0-5 client) and
immediately I am receiving a hector connection failure errors when drain
starts. With the hector connection issues, the external client stops sending
writes and it's also fails. Below show how I build up the connection with
hector.

Cluster cluster = HFactory.getOrCreateCluster(test-dev-cluster,
10.1.161.60,10.1.161.61,10.1.161.62);
Keyspace keyspace = HFactory.createKeyspace(getKeyValue(KEYSPACE_NAME_KEY),
cluster);

Other than the above 2 statements, I am passing any configuration to hector
to build up the connections. 

What I noticed is, every time hector uses zero element from the server list
url and still try to connect to the same server, if it fails. 

Could someone help me to solve this hector fail over?

Thanks

/Roshan 
 





--
View this message in context: 
http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/Cassandra-Hector-failover-load-balancing-not-as-expected-with-version-1-0-5-tp7581380.html
Sent from the cassandra-u...@incubator.apache.org mailing list archive at 
Nabble.com.