This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision | ||
admin:optimization:caches:filesystem [2012/02/07 17:47] doc |
admin:optimization:caches:filesystem [2016/06/28 22:38] (current) |
||
---|---|---|---|
Line 9: | Line 9: | ||
[[admin:optimization:caches:home|Overview]] | | [[admin:optimization:caches:home|Overview]] | | ||
- | [[admin:optimization:caches:hdc|Data Caching]] | | + | [[admin:optimization:caches:hdc|Hashed Data Caching]] | |
[[admin:optimization:caches:cachequal|Qualification Caching]] | | [[admin:optimization:caches:cachequal|Qualification Caching]] | | ||
**[[admin:optimization:caches:filesystem|Filesystem Caching]]** | **[[admin:optimization:caches:filesystem|Filesystem Caching]]** | ||
Line 17: | Line 17: | ||
==== Filesystem Caching ==== | ==== Filesystem Caching ==== | ||
- | Omnidex benefits greatly from caching by the filesystem, managed by the operating system. Omnidex purposefully maintains small index files, and purposefully uses standard file system calls, in order to insure the best caching by the operating system. This allows index files and temporary files to be as accessible as possible. | + | Omnidex benefits greatly from the filesystem caching that is automatically managed by the operating system. Omnidex purposefully maintains small index files and uses standard file system calls in order to insure the best caching by the operating system. |
- | To support this caching, it is important to maintain adequate physical memory on the server. Consult the [[install:requirements|System Requirements]] documentation for specific memory recommendations. In addition to adequate memory, it is important to avoid running processes on the Omnidex server that would flood the filesystem cache by doing excessive I/O. | + | To support filesystem caching, it is important to maintain adequate physical memory on the Omnidex server. Consult the [[install:requirements|System Requirements]] documentation for specific memory recommendations. In addition to adequate memory, it is also important to avoid running other processes on the Omnidex server that would flood the filesystem cache by doing excessive I/O. |
- | The applications most likely to flood the filesystem cache with excessive I/O are standard relational database, such as Oracle, SQL Server and MySQL. These databases use retrieval strategies that are much more dependent on direct access to the data, usually with multiple threads or processes. This will quickly flood the filesystem cache, and it will have a significant negative impact on Omnidex performance. | + | The applications most likely to flood the filesystem cache with excessive I/O are relational databases, such as Oracle, SQL Server and MySQL. These databases use retrieval strategies that are much more dependent on direct access to the data, usually with multiple threads or processes performing large amounts of I/O at the same time. This quickly floods the filesystem cache, and it will have a significant negative impact on Omnidex performance. |
- | When possible, it is preferable to run Omnidex on a server independent from the relational database. Typically, the database is then accessed through SQL*Net or ODBC. The database access is slower using this approach, but in the long run, this insures the best Omnidex performance. Additionally, since many Omnidex queries are processed solely in the indexes without accessing the database, this strengthens the recommendation to place Omnidex and the relational database on separate servers. | + | When possible, run Omnidex on a server that is separate from the relational database. In these situations the database will be accessed through the network, using protocols like SQL*Net or ODBC. This network interface may slow down Omnidex's access to the underlying database, but in the long run, this strategy insures the best overall Omnidex performance. This is especially true since Omnidex often optimizes queries so that no database access is required. Note that if Omnidex is accessing raw data files, then this separation is not needed. |
===== ===== | ===== ===== |