VIVO Documentation
Page History
...
Writing the MySQL transaction log
MySQL allows you to control its logging behavior, using the the innodb_flush_log_at_trx_commit parameter. On some systems, changing the value of this parameter can dramatically improve performance.
Using the default setting, the log is written to the file buffer and the buffer is flushed to disk at the end of each transaction. This is necessary to insure full ACID compliance, but the overhead is substantial. Most of VIVO is not transaction-oriented: each statement is added or deleted in its own transaction. So the default setting means that a physical write to disk is required for each new RDF statement.
Setting innodb_flush_log_at_trx_commit to 0 or 2 will greatly improve throughput, while adding a minimal level of risk to the data. Under some circumstances, with some settings, up to one second of transactions can be lost. Most VIVO installations will find this to be an acceptable level of risk.
| setting | meaning | worst case risk |
|---|---|---|
| 1 (default) | Write the log after each transaction. Flush to disk after each transaction. | If MySQL crashes, lose transactions in progress. On power failure or system crash, lose transactions in progress. |
| 2 | Write the log after each transaction. Flush to disk once per second. | If MySQL crashes, lose transactions in progress. On power failure or system crash, lose one second of transactions. |
| 0 | Write the log once per second. Flush to disk once per second. | If MySQL crashes, lose one second of transactions. On power failure or system crash, lose one second of transactions. |
This page provides full details regarding innodb_flush_log_at_trx_commit: http://dev.mysql.com/doc/refman/5.1/en/innodb-parameters.html#sysvar_innodb_flush_log_at_trx_commit
Setting the MySQL query cache size
Increasing the MySQL query cache size will likely translate into improved VIVO performance in that once large pages have been fetched once, they're typically quite a bit faster to load on later fetches.
Tracing back from SQL to SPARQL
...
Overview
Content Tools