voltdb

Run VoltDB stored procedures at regular interval from VoltDB


Is there any way to execute VoltDB stored procedures at regular interval or schedule store procedure to run at a specific time?

I am exploring VotlDB to shift out product from RDBMS to VotlDB. Out produce written in java.

Most of the query can be migrated into the VoltDB stored procedures. But In our product, we have cron job in oracle which executes at regular interval. Now I do not find such features in VoltDB.

I know VoltDB stored procedures can be called from the application at regular interval but our product deploys in an Active-Active mode, in that case, all application will call store procedure at regular interval and that is not a good solution or otherwise, we have to develop some mechanism to run procedure from one instance only.

so It would be good if I get cron job feature from VoltDB.


Solution

  • I work at VoltDB. There isn't currently a feature like this in VoltDB, for example like DBMS_JOB in Oracle.

    You could certainly use a cron job on one of the servers in your cluster, or on some other server within your network that could invoke sqlcmd to run a script or echo individual SQL statements or execute procedure commands through sqlcmd to the database. Making cron jobs highly available is a general problem. You might find these other discussions helpful:

    You could also look into something like rcron.

    One thing to be careful of when converting from an RDBMS to VoltDB is that VoltDB is optimized for processing many small transactions in parallel across many partitions. While the architecture of serialized execution per partition excels for many operational and streaming workloads, it is not designed to perform bulk operations on many rows at a time, especially transactions that need to perform writes on many rows that may be in different partitions within one transaction.

    If you have a periodic job that does something like "process all the new rows that meet some criteria" you may find this transaction is slow and every time it runs it could delay other parts of the workload, especially if many rows have accumulated. It would be more the "VoltDB Way" to replace a simple INSERT statement that you may be using to ingest data (to be processed later by a scheduled job) with a procedure that inserts and immediately processes the row of data. You might even need a procedure that checks for other records and processes small sets of rows as a group, for example stitching together segments of data that go together but may have arrived out of order. By operating on fewer records at a time within one partition at a time, this type of procedure would be more scalable and would keep the data closer to your desired finished state in real time, rather than always having some data waiting to be processed.