8.2 Infinite Cache
8 Performance Analysis and Tuning : 8.2 Infinite Cache
8.2 Infinite CacheNote: Infinite Cache has been deprecated and may be removed in a future release. Please contact your EnterpriseDB Account Manager or for more information.Advanced Server tries very hard to minimize disk I/O by keeping frequently used data in memory. When the first server process starts, it creates an in-memory data structure known as the buffer cache. The buffer cache is organized as a collection of 8K (8192 byte) pages: each page in the buffer cache corresponds to a page in some table or index. The buffer cache is shared between all processes servicing a given database.When you select a row from a table, Advanced Server reads the page that contains the row into the shared buffer cache. If there isn't enough free space in the cache, Advanced Server evicts some other page from the cache. If Advanced Server evicts a page that has been modified, that data is written back out to disk; otherwise, it is simply discarded. Index pages are cached in the shared buffer cache as well.A client application sends a query to the Postgres server and the server searches the shared buffer cache for the required data. If the requested data is found in the cache, the server immediately sends the data back to the client. If not, the server reads the page that holds the data into the shared buffer cache, evicting one or more pages if necessary. If the server decides to evict a page that has been modified, that page is written to disk.As you can see, a query will execute much faster if the required data is found in the shared buffer cache.One way to improve performance is to increase the amount of memory that you can devote to the shared buffer cache. However, most computers impose a strict limit on the amount of RAM that you can install. To help circumvent this limit, Infinite Cache lets you utilize memory from other computers connected to your network.With Infinite Cache properly configured, Advanced Server will dedicate a portion of the memory installed on each cache server as a secondary memory cache. When a client application sends a query to the server, the server first searches the shared buffer cache for the required data; if the requested data is not found in the cache, the server searches for the necessary page in one of the cache servers.When a client application sends a query to the server, the server searches the shared buffer cache for the required data. If the requested data is found in the cache, the server immediately sends the data back to the client. If not, the server sends a request for the page to a specific cache server; if the cache server holds a copy of the page it sends the data back to the server and the server copies the page into the shared buffer cache. If the required page is not found in the primary cache (the shared buffer cache) or in the secondary cache (the cloud of cache servers), Advanced Server must read the page from disk. Infinite Cache improves performance by utilizing RAM from other computers on your network in order to avoid reading frequently accessed data from disk.You can add or remove cache servers without restarting the database server by adding or deleting cache nodes from the list defined in the edb_icache_servers configuration parameter. For more information about changing the configuration parameter, see Section 18.104.22.168.When you add one or more cache nodes, the server re-allocates the cache, dividing the cache evenly amongst the servers; each of the existing cache servers loses a percentage of the information that they have cached. You can calculate the percentage of the cache that remains valid with the following formula:(existing_nodes * 100) / (existing_nodes + new_nodes)For example, if an Advanced Server installation with three existing cache nodes adds an additional cache node, 75% of the existing cache remains valid after the reconfiguration.If cache nodes are removed from a server, the data that has been stored on the remaining cache nodes is preserved. If one cache server is removed from a set of five cache servers, Advanced Server preserves the 80% of the distributed cache that is stored on the four remaining cache nodes.When you change the cache server configuration (by adding or removing cache servers), the portion of the cache configuration that is preserved is not re-written unless the cache is completely re-warmed using the edb_icache_warm() function or edb_icache_warm utility. If you do not re-warm the cache servers, new cache servers will accrue cache data as queries are performed on the server.Without Infinite Cache, Advanced Server will read each page from disk as an 8K chunk; when a page resides in the shared buffer cache, it consumes 8K of RAM. With Infinite Cache, Postgres can compress each page before sending it to a cache server. A compressed page can take significantly less room in the secondary cache, making more space available for other data and effectively increasing the size of the cache. A compressed page consumes less network bandwidth as well, decreasing the amount of time required to retrieve a page from the secondary cache.The fact that Infinite Cache can compress each page may make it attractive to configure a secondary cache server on the same computer that runs your Postgres server. If, for example, your computer is configured with 6GB of RAM, you may want to allocate a smaller amount (say 1GB) for the primary cache (the shared buffer cache) and a larger amount (4GB) to the secondary cache (Infinite Cache), reserving 1GB for the operating system. Since the secondary cache resides on the same computer, there is very little overhead involved in moving data between the primary and secondary cache. All data stored in the Infinite Cache is compressed so the secondary cache can hold many more pages than would fit into the (uncompressed) shared buffer cache. If you had allocated 5GB to the shared buffer cache, the cache could hold no more than 65000 pages (approximately). By assigning 4GB of memory to Infinite Cache, the cache may be able to hold 130000 pages (at 2x compression), 195000 pages (at 3x compression) or more. The compression factor that you achieve is determined by the amount of redundancy in the data itself and the edb_icache_compression_level parameter.To use Infinite Cache, you must specify a list of one or more cache servers (computers on your network) and start the edb_icache daemon on each of those servers.Please Note: Infinite Cache and the effective_io_concurrency parameter can potentially interfere with each other. You should disable asynchronous I/O requests (by setting the value of effective_io_concurrency to 0 in the postgresql.conf file) if you enable the Infinite Cache feature.
8 Performance Analysis and Tuning : 8.2 Infinite Cache