Approach to KV queries
Aug 8, 2010
I think I've got a basic understanding as to how a VoltDB Transaction Manager will query other VoltDB nodes as necessary in order to appropriately respond to a client request (via stored procedures). My question relates to the way in which VoltDB handles key-value queries. Will a KV query go directly to the VoltDB node that has the data as in a typical distributed hash table store, or does the query go to any VoltDB node, with the risk that the node that receives the request may not have the data and may need to relay data from the node that has the KV data?
I'm looking forward to running my own latency tests as soon as I get my test network built, but I'm curious as to whether using KV is a reasonable way to reduce overall latency for certain latency-sensitive queries. My application probably can't support a clone of the necessary data at each node, and the ~10ms latency cited for SQL queries is too slow for part of my application (it will work for the rest, hence my question about mixed SQL and KV).
Looking forward to the KV webinar later in the month.
Cheers,
Dean
I'm looking forward to running my own latency tests as soon as I get my test network built, but I'm curious as to whether using KV is a reasonable way to reduce overall latency for certain latency-sensitive queries. My application probably can't support a clone of the necessary data at each node, and the ~10ms latency cited for SQL queries is too slow for part of my application (it will work for the rest, hence my question about mixed SQL and KV).
Looking forward to the KV webinar later in the month.
Cheers,
Dean