Pools

Destination hosts are grouped into "pools". A pool is a basic building block of a routing config. At a minimum, a pool consists of an ordered list of destination hosts and a hash function.

Hash functions

• "ch3" ("consistent hash, version 3") is the consistent hashing algorithm used by mcrouter. This is the default hash function for pools. The underlying algorithm is furc_hash(). The routing part of the key is hashed and the request is set to the resulting index in the pool.
• "crc32" hash sends the request to host index crc32(routing_key) % pool_size.
• HostIdRoute doesn't look at the key at all, instead always sending to hostid % pool_size, where hostid is some unique index associated with mcrouter's local host. As a result, a single client will always talk to the same HostIdRoute destination, but two different clients will likely talk to different destinations.

Shadowing

Mcrouter supports shadowing production traffic. Shadowing works on the pool level. A pool (called "production pool" in the following discussion) can have multiple shadow pools; each shadow is defined with an index_range, which is a continuous range of production host indices; and a key_fraction_range which is a subinterval of [0.0, 1.0] and used for selecting a fraction of the total key space. Only requests which would go to one of the production hosts in the index_range with keys which would hash in the key_fraction_range (the hash used for this purpose is SpookyHashV2) are sent to both the production and the shadow hosts.