github.com/outbrain/consul@v1.4.5/website/source/intro/vs/zookeeper.html.md (about) 1 --- 2 layout: "intro" 3 page_title: "Consul vs. ZooKeeper, doozerd, etcd" 4 sidebar_current: "vs-other-zk" 5 description: |- 6 ZooKeeper, doozerd, and etcd are all similar in their architecture. All three have server nodes that require a quorum of nodes to operate (usually a simple majority). They are strongly-consistent and expose various primitives that can be used through client libraries within applications to build complex distributed systems. 7 --- 8 9 # Consul vs. ZooKeeper, doozerd, etcd 10 11 ZooKeeper, doozerd, and etcd are all similar in their architecture. All three have 12 server nodes that require a quorum of nodes to operate (usually a simple majority). 13 They are strongly-consistent and expose various primitives that can be used through 14 client libraries within applications to build complex distributed systems. 15 16 Consul also uses server nodes within a single datacenter. In each datacenter, Consul 17 servers require a quorum to operate and provide strong consistency. However, Consul 18 has native support for multiple datacenters as well as a more feature-rich gossip 19 system that links server nodes and clients. 20 21 All of these systems have roughly the same semantics when providing key/value storage: 22 reads are strongly consistent and availability is sacrificed for consistency in the 23 face of a network partition. However, the differences become more apparent when these 24 systems are used for advanced cases. 25 26 The semantics provided by these systems are attractive for building service discovery 27 systems, but it's important to stress that these features must be built. ZooKeeper et 28 al. provide only a primitive K/V store and require that application developers build 29 their own system to provide service discovery. Consul, by contrast, provides an 30 opinionated framework for service discovery and eliminates the guess-work and 31 development effort. Clients simply register services and then perform discovery using 32 a DNS or HTTP interface. Other systems require a home-rolled solution. 33 34 A compelling service discovery framework must incorporate health checking and the 35 possibility of failures as well. It is not useful to know that Node A provides the Foo 36 service if that node has failed or the service crashed. Naive systems make use of 37 heartbeating, using periodic updates and TTLs. These schemes require work linear 38 to the number of nodes and place the demand on a fixed number of servers. Additionally, 39 the failure detection window is at least as long as the TTL. 40 41 ZooKeeper provides ephemeral nodes which are K/V entries that are removed when a client 42 disconnects. These are more sophisticated than a heartbeat system but still have 43 inherent scalability issues and add client-side complexity. All clients must maintain 44 active connections to the ZooKeeper servers and perform keep-alives. Additionally, this 45 requires "thick clients" which are difficult to write and often result in debugging 46 challenges. 47 48 Consul uses a very different architecture for health checking. Instead of only having 49 server nodes, Consul clients run on every node in the cluster. These clients are part 50 of a [gossip pool](/docs/internals/gossip.html) which serves several functions, 51 including distributed health checking. The gossip protocol implements an efficient 52 failure detector that can scale to clusters of any size without concentrating the work 53 on any select group of servers. The clients also enable a much richer set of health 54 checks to be run locally, whereas ZooKeeper ephemeral nodes are a very primitive check 55 of liveness. With Consul, clients can check that a web server is returning 200 status 56 codes, that memory utilization is not critical, that there is sufficient disk space, 57 etc. The Consul clients expose a simple HTTP interface and avoid exposing the complexity 58 of the system to clients in the same way as ZooKeeper. 59 60 Consul provides first-class support for service discovery, health checking, 61 K/V storage, and multiple datacenters. To support anything more than simple K/V storage, 62 all these other systems require additional tools and libraries to be built on 63 top. By using client nodes, Consul provides a simple API that only requires thin clients. 64 Additionally, the API can be avoided entirely by using configuration files and the 65 DNS interface to have a complete service discovery solution with no development at all.