...
The redundancy_factor (rf) is a number [1, 100] that determines the redundancy pool (rp) or the number of nodes which data for a particular key/locator pair are stored. This is the exact number of nodes that are eligible to respond to a query for a read that touches a key in a record. The redundancy pool is a subset of the partition pool.
rp = 1 or ceil(rf * pp)
Distribution
| 0 | 1 | 2 | 3 | 4 | 5 |
|---|
Consider a cluster that has 6 nodes. We're also making a few assumptions that are worth calling out:
- The server prefs file contains an order (comma separated) list of node addresses
- The server reads that list of nodes on startup and caches the list somewhere on disk
- If the server starts up and notices the list of nodes in the prefs file is different from the one it cached, then it needs to do a rebalance!
Locator Hashing
| Code Block |
|---|
// We need to get pp distinct nodes for the pool, so we continually hash the hash of the locator until we have everything we need.
Integer[] partitionPool = new Integer[pp]
Integer hash = null;
for(int i = 0; i < pp ; i++){
int slot = -1;
while(slot == -1 || partitionPool[slot] != null){
slot = hash(hash == null ? locator : hash) % n; //n is number of nodes
}
partitionPool[i] = slot;
}
// We now have an array of node identifiers that form our partition pool. |
Locator and Key Hashing
| Code Block |
|---|
// We need to get rp nodes from the partition pool, so we continually hash the hash of the locator and key until we have everything we need
Integer[] redundancyPool = new Integer[rp]
Integer hash = null;
for(int i = 0; i < rp; i++){
int slot = -1;
while(slot == -1 || redundancyPool[slot] != null){
slot = hash(hash == null ? locator + key : hash) % partitionPool.length;
}
redundancyPool[i] = partitionPool[slot];
}
//We now have an array of node identifiers that for our redundancy pool. |