| Column Name | Mode | Description | Data Type |
|---|---|---|---|
| RRDHCount | count | Total count of hashed row redistributions (RRDs) processed by the node-level buffered redistribution services (NLB). The RRDHCasc[All|Most|Some] fields count subsets of those RRDs based on the percentage of AMPs in the RRD cascade tree where the rows do not overflow the cascade buffer but are fully contained within the cascade rather than being sent directly to their targeted destination AMPs by NLB's minicast S2S-Redist messages. This percentage is referred to as the "Cascade Factor" and varies between 0% where the RRD rows overflow the cascade buffer in every AMP, and 100% where the rows are cascaded all the way to the cascade root AMP and do not overflow the cascade buffer even there. | FLOAT |
| RRDHCascAll | count | Count of hashed RRDs with a Cascade Factor of 100%. In this case all rows are cascaded up to the cascade root AMP without overflowing the cascade buffer anywhere, including in the cascade root itself. | FLOAT |
| RRDHCascMost | count | Count of hashed RRDs with a Cascade Factor of 90% or more, but less than 100%. In this case the rows are cascaded without a buffer overflow in most but not all AMPs, where "most" is based on a 90% super-majority, not a simple majority in the usual sense of anything over 50%. | FLOAT |
| RRDHCascSome | count | Count of hashed RRDs with a Cascade Factor greater than 0% but less than 90%. In this case the rows are cascaded without a buffer overflow in some AMPs, even as few as one, but less than "most' AMPs as defined by the 90% super-majority. Note that there is no field (e.g., "RRDHCascNone") that counts the hashed RRDs with a Cascade Factor of exactly 0%, where the rows overflow the cascade buffer in every AMP. However, this virtual count can be computed from the other fields as follows: RRDHCascNone = RRDHCount - (RRDHCascAll + RRDHCascMost + RRDHCascSome)
|
FLOAT |