Template


1	Models	HDFS	Cassandra	Redis	MongoDB
2	Clustering	Master	Masterless	Masterless	Multi-master
3	Sharding	Policy-based	Range-based	Range-based	Policy-based
4	Mutability	Out-of-place	Temporary out-of-place (In-place)	Direct update (in-place)
5	Clustering Mgr Service
6	Membership tracking	Centralized	Gossip	Gossip	Central + Distributed
7	IO completion with offline replicas	Abort	Continue till quorum	Always continue	Continue till quorum
8	Offline replica update	NA	Hinted Handoff	NA
9	Tracking replica lag	Always in-sync	None	None
10	Consensus Mgt Service
11	Consensus protocol	Leader-based	Quorum-based	None	Leader-based
12	Leader selection protocol	Heuristic	Heuristic	None	Heuristic
13	Cluster-wide HA Service
14	Data Node failover	Selected node (Master)	Sharded Neighbor	Sharded Neighbor	Selected node (Master)
15	Metadata node failover	Failover to backup	Collocated w/ Data	Collocated w/ Data	Clustered MD
16	Partition handling	Unavailable (CP)	Cluster-based quorum	Cluster-based quorum	Unavailable (CP)
17	Namespace Lookup Service
18	Cluster-wide metadata representation	Name-to-resource	Distributed Hashtable	Distributed Hashtable	Name-to-Obj Mgr
19	Cluster-wide metadata freshness	Materialized + Derived	Fully derived	Fully derived	Materialized + Derived
20	Object lookup	Client-driven	Cluster-driven	Client-driven	Cluster-driven
21	Replica lookup	Unlisted	Listed	Listed	Listed
22	Consistency Service
23	Mutal Exclusion Protocol	Non-blocking	Blocking	Non-blocking	Blocking
24	Atomicity Strategy	Single Object - Non-blocking	Single Object - Non-blocking	Single Object- Blocking	Single Object- Blocking
25	Ordering	Single Object - Non-blocking	Single Object - Non-blocking	Single Object- Blocking	Multi-object
26	Transactional support	None	ME + Atomicity	None
27	Durability Service
28	Durability model	Replication	Replication	Replication	Replication
29	Replica update model	Async	Async	Async	Async
30	Replica byte-level fidelity	Atleast once w/ ordering	Atleast once wo/ ordering	Atleast once w/ ordering	Atleast once w/ ordering
31	Replica consistency	Eventual: Monotonic	Eventual: Read-your-writes	Weak	Eventual: Monotonic
32	Propogating replica updates	State-based	State-based	Update history	Ops shipping
33	Selecting replica coordinator for updates	Delegated primary	Sharded	Sharded	Delegated primary
34	Ordering replica updates	Ordered	Unordered	Unordered	Ordered
35	Replica selection during reads	Not Applicable	Fastest
36	Replica placement	Policy	Sharding	Heuristics
37	Replica access	Not accessible	Any replica Read/Write	Any Replica Reads
38	Geo-redundancy protocol	None	Sync + async replicas
39	Object Manager Service
40	Resource Unit	File	File	Mem Address	File
41	Namespace-to-resource unit	1: Many Reosurce Unit	Many:1 Resource Unit	1:1	Many:Many
42	Per-object Metadata persistence	Separate	Collocated	Collocated	Collocated
43	GC schedule	Release -- Return to Free Pool	Major/Minor compact	Release -- Return to Free Pool	Major/Minor compact
44	Delete model	Delete Data and Metadata	Delete Data, Tombstome Metadata	Delete Data and Metadata	Delete Data and Metadata
45	Buffering/Caching
46	Buffer Flushing Model	Explicit Window	Implicit control	None	Implicit control
47	Buffering RW Coherence	Using leases		None	Using leases
48	Placement of data caching service	Client	Data Node
49	Data Caching Strategy	On access	None
50	Cache Invalidation Protocol	On error	None
51	Resource Management
52	On-disk records	Schema	Data Structure	Data Structure	Data Structure
53	Striping across physical resources	Across nodes	Within node	None	Within node
54	Managing HW heterogenity	None	Propotional allocation	None	Propotional allocation