Hi Teradata Experts,
I'm new to Teradata. I'm basically an Oracle DB Architect. I got a nice opportunity to work on one of an existing Teradata DW Application in our Organization.
I would like to ask a question.
In Oracle, when a user wants to establish a session with the Server, the Oracle Listener will receive the request and push that request to any one of the Oracle Instance either based on Round Robin/Based on the Instances Work Load. From there on, the Instance handles the session and if something happens to that Instance and it crashed, the session will be moved to another instance and that is transparent to the user.
In Teradata i see each and every node has some PE and AMPS and Each AMP manages its own data and data is not shared between other AMPS.
When a user tries to establish a connection to Teradata, will Teradata Gateway receive and send that request to any 1 NODE?(if yes, on what basis)
Thanks a lot for your guidance. I understood the concept behind LCC.
THis bring another question in my mind. could you pls throw some light here too please.
assume my session got connected to node 1 (assume i have 4 node MPP) and i'm writing a query which my PE in Node1 parses and identifies the data is in Node 3's AMP5. In this case, will the PE in Node1 will automately route the request to the MAP5 in Node3 thru BYNET? is My Understanding Right? or the PE in Node1 will talk only to AMP in NODE1 ?
PEs can communicate with any/all AMPs as needed, regardless of whether the AMP is located on the same node as the PE ("on-node") or a different node ("off-node"). The Bynet handles all the messaging. Generally speaking, applications do not need to be concerned with this.
Regarding your original question: "When a user tries to establish a connection to Teradata, will Teradata Gateway receive and send that request to any 1 NODE?(if yes, on what basis)"
Sessions are owned by PEs. At logon time, the Teradata Database Gateway chooses a PE for the session. Typically, the Teradata Database Gateway will choose the PE with the fewest sessions. The chosen PE may be either on-node or off-node with respect to the Gateway.
Adding to the excellent explanation by mr. Nolan, you can have more than one PE per node (as a matter of fact, this is the usual case). Same with gateways. Both are vprocs (software implementations).