If you’ve used memcached before, you probably know that you can add data to the end of an existing string with APPEND. The documentation for memcached states that APPENDcan be used for managing lists of items. Great! You add items to the end of the string you’re treating as a list. But then how do you remove items? The memcached answer is to use a blacklist to hide items, in order to avoid read/update/write (or a database query and memcached write). In Redis, you’d either use a LISTor a SETand then add and remove items directly.
By using Redis instead of memcached for this and other problems, not only can your code be shorter, easier to understand, and easier to maintain, but it’s faster (because you don’t need to read a database to update your data). You’ll see that there are also many other cases where Redis is more efficient and/or easier to use than relational databases.
One common use of databases is to store long-term reporting data as aggregates over fixed time ranges. To collect these aggregates, it’s not uncommon to insert rows into a reporting table and then later to scan over those rows to collect the aggregates, which then update existing rows in an aggregation table. Rows are inserted because, for most databases, inserting rows is a very fast operation (inserts write to the end of an on-disk file, not unlike Redis’s append-only log). But updating an existing row in a table is fairly slow (it can cause a random read and may cause a random write). In Redis, you’d calculate the aggregates directly using one of the atomic INCRcommands — random writes to Redis data are always fast, because data is always in memory, 2 and queries to Redis don’t need to go through a typical query parser/optimizer.
By using Redis instead of a relational or other primarily on-disk database, you can avoid writing unnecessary temporary data, avoid needing to scan over and delete this temporary data, and ultimately improve performance. These are both simple examples, but they demonstrate how your choice of tool can greatly affect the way you solve your problems.
As you continue to read about Redis, try to remember that almost everything that we do is an attempt to solve a problem in real time (except for task queues in chapter 6). I show techniques and provide working code for helping you remove bottlenecks, simplify your code, collect data, distribute data, build utilities, and, overall, to make your task of building software easier. When done right, your software can even scale to levels that would make other users of so-called web-scale technologies blush.
We could keep talking about what Redis has, what it can do, or even why. Or I could show you. In the next section, we’ll discuss the structures available in Redis, what they can do, and some of the commands used to access them.
2 To be fair, memcached could also be used in this simple scenario, but with Redis, your aggregates can be placed in structures that keep associated aggregates together for easy access; the aggregates can be a part of a sorted sequence of aggregates for keeping a toplist in real time; and the aggregates can be integer or floating point.