| Thread Previous • Date Previous • Date Next • Thread Next |
On 1/3/13 7:07 AM, Leander Bessa Beernaert wrote:
Hello all, Could some please confirm my analysis of the ring creation in swift? I need to run a few benchmarks on a set of data (200GB) total and need it to be replicated at least 3 times. From the Swift docs [1], I understood that the command below create the ring for a partition sized 2^18 and 3 replicas. swift-ring-builder account.builder create 18 3 1 # swift-ring-builder container.builder create 18 3 1 # swift-ring-builder object.builder create 18 3 1 Now, in my case i should use the values 28 3 1 right, or should i factor in the replication and instead use 30 3 1?
Neither one. The partitions in the Swift ring are subsets of the range of a hash function; they don't have anything to do with the number of bytes of storage in your cluster.
General rule of thumb is to figure out how many drives your cluster will ever have at its largest, and then choose the smallest part_power that gives you ~100 partitions per drive.
I have at the moment 1 proxy node and 5 storage nodes with a 200GB storage partition. This won't be enough to accomodate everything right?
I don't see why not. You've got 200GB stored data * 3 replicas = 600GB raw data used, and you've got 1000GB of disk. Even with a little bit of overhead for accounts and containers, that's still plenty of disk.
| Thread Previous • Date Previous • Date Next • Thread Next |
