drupal vm memcached mixing up 2 different sites

What this means in practice is that GET and SET calls maintain a constant speed as your cache size grows (this is a very good thing!). They are both designed to provide O(1) performance for all GET and SET operations. Redis and Memcached both provide a similar service: an in-memory key/value store. (APC is obviously still very usable as an opcode cache, even when you have multiple servers.)

APC in particular is unable to share its cache across multiple servers, which makes it unusable as a data cache for any site that uses more than one web server, and even on a single server the user cache suffers from fragmentation and does not implement the least recently used (LRU) eviction. There are other cache storage options, such as the APC user cache (this allows you to use the APC cache for objects other than PHP opcode) however, Redis and Memcache are by far the most widely used options for large Drupal sites, as they perform well and are generally much easier to scale. So far we have concentrated on two alternatives: Redis and Memcached. This is why even though replacing the cache backend may not improve speed for an individual request, it can actually make a dramatic difference for large or heavily loaded sites. Placing data into a key/value store, which will maintain performance as the data size increases to multiple gigabytes and allows for easy horizontal scaling Offloading queries from MySQL, allowing the server to perform better for other queries By swapping out cache backends, we achieve two things: Having your caching layer impacting the speed of your primary data store is not really acceptable. MySQL, on the other hand, is far more difficult to scale horizontally and is likely to also be being used as your primary data source. Key/value stores such as Memcached and Redis are great at dealing with large amounts of data, and Memcached in particular can be scaled horizontally simply by adding more servers. The true benefit becomes clearer as traffic to the site increases. For cache gets, this is often not the case-a well-tuned MySQL server can perform as well as or better than Memcache in some cases, although Memcache should almost always outperform MySQL for cache sets. One incorrect assumption that many people make is that replacing the cache backend will automatically make each individual cache request faster. Swapping out the cache backend can help improve both performance and scalability however, we should look closely at exactly what changes when the cache storage is replaced. Not only can these backends help to offload queries from MySQL, but they can be easily scaled out horizontally as your site grows.Īs introduced in Chapter 3, Drupal’s cache API stores data in MySQL by default, but it can be easily integrated with other technologies. There are multiple Drupal contributed projects that provide a way to pull cache entries out of MySQL and store them in an alternative storage backend, such as Memcached or Redis. This can cause increased network traffic and ultimately increase the load on the database servers.

While this is generally fine for smaller sites, as the number of cache and session objects grows, frequent actions such as cache invalidation put a high demand on the database layer. Queries with a condition on one field that sort on another then have to join across multiple tables, which is very hard to optimize with a relational database.Īnother common performance bottleneck that particularly affects higher-traffic Drupal sites is that cache and session data are stored in MySQL tables by default. The entity/field storage provided by core provides one or two tables for each core entity, then two tables for each individual field (one for the current or “live” revision and one for older revisions). This is a very powerful set of features, but it can make getting great performance out of a traditional relational database quite difficult. Entity types such as nodes, users, and comments can be extended with configurable fields, and then the Views module allows those to be queried and presented-all without leaving the user interface. Drupal is very modular and flexible in terms of how data can be defined. MySQL provides a great database backend for Drupal, but there are certain issues that arise, especially as Drupal sites get larger. This includes low-level APIs such as caching and queues, as well as site content. Chapter 16. Alternative Storage and Cache Backendsĭrupal uses MySQL as the default storage backend for all data on a site.

Drupal vm memcached mixing up 2 different sites