While each user has 130 friends on average, they have approximately 48000 friends-of-friends5. Even assuming only 4 bytes per posting list entry, this would still take over 178TB to store the posting lists for 1 billion users, which would require hundreds of machines.
Catalog
Quotes tagged ‘Understanding Queries’
Jon is also a friend-of-friends of himself since he is—by definition—a friend of his friends.
For storing per-entity metadata, Unicorn provides a forward index, which is simply a map of id to a blob that contains metadata for the id. The forward index for an index shard only contains entries for the ids that reside on that shard.
Unicorn also supports a Difference operator, which re-turns results from the first operand that are not present in the second operand. Continuing with the example above, we could find female friends of Jon Jones who are not friends of Lea Lin by using (difference (and friend:5 gender:1) friend:6). A…
Although there are many billions of nodes in the social graph, it is quite sparse: a typical node will have less than one thousand edges connecting it to other nodes. The average user has approximately 130 friends. The most popular pages and applications have tens of millions of edges, but these pag…
Result set scoring offers yet another layer of filtering that looks at a number of entities together and returns a subset of these entities that are most interesting as a set (and not necessarily the highest scoring set of results).
Here 273819889375819 is the fbid of the category ‘restaurants,’ and 20531316728 is the fbid of Facebook. So the query says ‘intersect places of category restaurants with places liked by employees of Facebook.
Unicorn is designed to answer billions of queries per day at latencies in the hundreds of milliseconds, and it serves as an infrastructural building block for Facebook’s Graph Search product.
1. Come up with an idea for a ranking change – ideas come from a mix of creativity on the part of our ranking engineers and feedback from our users 2. Implement the ranking change, test it, and launch it to a very small fraction of our user base 3. Measure the impact of the ranking change
It is the richness of the data that defines the nature of Graph Search; the system needs to be designed toward understanding the user intent precisely and serving structured objects.
In Graph Search, we have 20+ entity categories, including {user}, {group}, {application}, {city}, {college}, etc. At entity detection time, we allow multiple query segments, including overlapping ones, to be detected as potential entities, and allow multiple entity categories to be assigned to each…