Comparisons of no sql databases march 2014
- 1. Desc Score WS Database type 2% Key-Value, document database, JSON, schemaless, high availability in- memory(Caching) asynchronous persistence and cache model formed as a merger of Apache Couch DB and memcache. Does provide a mobile DB synced with original DB-no other DB has it. Has inmemory solution which Mongo and Accumulo do not have. 99 1.98 Best used 2% Any application where low- latency data access, high concurrency support and high availability is a requirement. Can be used as OLTP for inmemory transactions. 99 1.98 Use Cases and Adaptability(specific to ADP) 2% Low-latency use-cases like ad targeting or highly-concurrent web apps like online gaming (e.g. Zynga). 66 1.32 Storage Type 2% Key-Value with inmemory document datastore 99 1.98 General Key Facts - 10% Metrics Sub Metric Weighted Score CouchBase
- 2. Characteristics 2% very flexible but rather slow indexes 66 1.32 Data Storage 2% Volatile memory, File System 99 1.98 Unicode 1% Yes 99 1.98 Search Integration 2% External Plug-in 66 1.32 Compression 1% Yes 99 1.98 Conditional Entry updates 1% Yes 99 1.98 TTL for Entries 1% Yes 99 1.98 Graph support 2% No 0 0 Rich Design & Features - 10%
- 3. Query Language 2% JavaScript, Memcached-protocol Gartner's survey reports difficulties integrating with other DBMS. 33 0.66 Programming language 2% C, C++, Erlang. More language support is needed. 33 0.66 Ease of use(JSON) 2% Yes 66 1.32 Protocol Used 2% memcached + extensions 66 1.32 MapReduce 2% Yes 66 1.32 Integrity model 2% MVCC. No SPOF 99 1.98 R & D Velocity Acceleration - 10% Integrity - 10%
- 4. ACID transactions 5% Couchbase claims to be ACID- compliant on a per-item basis, but has no multi-operation transactions. Couchbase clients connect to a server list (or via a proxy) where keys are sharded across the nodes. Couchbase nodes inherit memcachedās default (and recommended) connection limit of 10k. ACID(Atomicity - Y, Consistency - Y, Isolation - Y, Durability - Y). 99 1.98 Transactions 1% No. Transactions are ACID at document level. 33 0.66 Referential Integrity 1% No 33 0.66 Revision Control 1% Yes 99 1.98 Secondary Indexes 4% Yes 99 1.98 Composite keys 2% Yes 99 1.98 Full text search 2% Yes 99 1.98 Throughput 2% Better(Need number) 66 1.32 In-Memory 1% Memcache is in-memory KV store 99 1.98 Geospatial Indexes 1% Yes 99 1.98 Performance - 15% Integrity - 10%
- 5. Relabalancing 1% Initaite manually but cluster is running and servicing requests 66 1.32 Latency 2% low latency 66 1.32 Replication Architecture 5% Multi-master replication and replica sets. Couchbase supports two types of replication. For intra- datacenter clusters, Couchbase uses membase-style replication, which favors immediate consistency in the face of a network partition. For multi- datacenter deployments, CouchDBās master-master replication is used. 99 1.98 Horizontal Scaling / Sharding(Share Nothing) Scoring: Auto Shard, Shard Manually, No Sharding 10% Yes. Autosharding. Hash. 99 1.98 Operating System 3% No Support yet for SuSE 66 1.32 Performance - 15% Infrastructure Scaling - 15%
- 6. Mangement / Monitoring GUI 3% GUI and CLI 99 1.98 Documentation 3% Good 66 1.32 Backup / Recovery 3% Not real time 33 0.66 Engineering & Installation 3% Easy 99 1.98 Cost and ROI 3% Reasonable Price point 33 0.66 Customer Base 3% 350 Customers total. Over 9500 paid servers are in use by several indutries veritical. 66 1.32 License 3% Apache(Community edition), Proprietary(Enterprise edition) 66 1.32 Professional Support 3% Evolving 66 1.32 Operational Adaptability - 15% Cost and Market Direction - 15%
- 7. Technology Depth & Competition 3% Market depth is limited to Key Value database. Huge competition is mounting from MongoDB 33 0.66 Total Score 100% 63 Cost and Market Direction - 15%
- 8. Desc Score WS Desc Score WS Desc Score Basic unit of organization is document storage, encoded in JSON, XML, Text or binary. Everything is compressed into binary trees based on Xpath Data model technique. 99 1.98 Document database, schemaless using BSON(and added JSON later). Used by ADP for mobile solution across 17 countries for mm+ customers. Trying to introduce search functionality 99 1.98 Leverages the Oracle Berkeley DB Java Edition High Availability storage engine to provide distributed, highly-available key/value storage for large- volume, latency-sensitive applications or web services. 99 It is a document-centric, transactional, search-centric, structure-aware, schema- agnostic, XQuery- and XSLT- driven, high performance, clustered, database server. 66 1.32 If you prefer to define indexes, not map/reduce functions. Cannot be used for OLTP. Good for document storage and retrieval not for almost realtime applications. Scaling becomes complex. 66 1.32 Provides fast, reliable, distributed storage to applications that need to integrate with ETL processing. 66 Government, Publishing, finance and many other large- scale sectors such as Medicare and Medicaid services, Dow Jones, Federal Aviation Administration. 66 1.32 can easily replace RDBMS with no schema so faster and no predefined columns, good for datastore, CRM applications 99 1.98 Social networks, Online retail, Web applications, Backup services for mobile devices. 99 Document stores, Native XML DBMS. 66 1.32 Document 66 1.32 Distributed Key-Value store 66 MarkLogic MongoDB Oracle NoSQL
- 9. Role-based security features JSON Storage Direct use of HDFS Multiple indexing strategies ACID Consistency Kerberos/LDAP support 66 1.32 Consistency Partition Tolerance Persistence 99 1.98 No single point of failure Multi-Node backup Optimized Hardware (Oracle Big Data appliance) Predictable latency 99 Native XML DBMS, Documents stored as compressed binary trees. 66 1.32 Memory Mapped files 66 1.32 Stored as key-value pairs, which are written to particular storage node(s), based on the hashed value of the primary key. 66 Yes 99 1.98 yes 99 1.98 Yes 99 Search includes many features listed in the comment Although many features are there, Solr / Elastic Search integration is still an involved exercise 66 1.32 Building search capability. MongoDB has a drive to integrate with Elastic Search 66 1.32 No 33 Data is stored as compressed binary trees. 99 1.98 yes 99 1.98 No (Need to clarify) 33 Yes 99 1.98 Yes 99 1.98 Yes 99 Yes ( Need to clarify) 99 1.98 Yes 99 1.98 Yes (Need to clarify) 99 Yes (Supports for semantics in that MarkLogic can store RDF triples, using SPARQL as its query language.) 66 1.32 No 0 0 Yes (RDF Graphs) 66
- 10. Xquery, JSON, Java API, REST, XML 99 1.98 API calls, JavaScript, Rest. Hadoop Connectior to and from HDFS. 99 1.98 Java/C API 66 C++ 66 1.32 C++ 66 1.32 Java 66 JSON 66 1.32 Better handling of documents, collections 99 1.98 Yes 99 XDMP (X Display Manager Protocol) 66 1.32 Custom, binary (BSON) 66 1.32 TCP(RMI), TCP(Proprietary) 66 Can use C++ to do Map/Reduce functions/calculations. 33 0.66 yes 66 1.32 Can use MapReduce when integrated with Hadoop environment 66 ACID, MVCC, No single point of failure 99 1.98 Not MVCC but you can sepratey use Mongo MVCC 99 1.98 ACID 99
- 11. Yes, need more information on what transactions are included. ACID. 99 1.98 MongoDB does not support multi-document transactions. However, MongoDB does provide atomic operations on a single document. D ( A - Conditional, C - Yes, Two phased commit is required. Uses memory mapped files for data storage, I - N) 99 1.98 Provides ACID complaint transactions for full Create, Read, Update and Delete (CRUD) operations, with adjustable durability and consistency transactional guarantees. ACID. 99 Yes 66 1.32 No. Transactions are ACID only at document level. 33 0.66 Yes 99 Yes ( Need to clarify) 66 1.32 No 33 0.66 No 33 Yes 99 1.98 Yes 99 1.98 Yes 99 Yes 99 1.98 Yes 99 1.98 Yes 99 Yes 99 1.98 Yes 99 1.98 Yes 99 Yes 99 1.98 No 99 1.98 No 0 Throughput is average 66 1.32 OK(Need number) 33 0.66 Better (Need number)_ 33 In-Memory stands can be configured 66 1.32 Memory Mapped files 66 1.32 Not in-memory 33 Yes 66 1.32 Yes 99 1.98 No (Need verification) 33
- 12. Yes 33 0.66 Initiate manually but cluster needs to be pulled down 33 0.66 Automatic Rebalancing 66 Average Latency ranges about 1.2ms 66 1.32 high at >20k 66 1.32 Low latency 66 Flexible Replication (Maintains copies of data on multiple servers. Original content is created by an application on master server. Replication copies that content to one or more replicas. Master and replicas are in different clusters which may or may not be in same location. It is asynchronous. Not a multi- master replication as documents updated by each application must be in different domains or this may cause unpredictable behavior due to overlap.) 66 1.32 Master-Slave-Replication for more than 12 nodes, Replica set is the preferred method, need arbiters or a separate machine and odd number for replication 66 1.32 Master-Slave Replication 66 Yes. Distributed architecture makes it easy to scale. 99 1.98 Yes. Scale Manually. Hash & Range. 66 1.32 Yes. Autosharding. 99 Windows, Solaris, Linux, OS X 99 1.98 Solaris, Linux, Windows, Mac OS X 99 1.98 Linux,OS X, Windows 99
- 13. Administration GUI 66 1.32 Monitoring GUI than Management GUI 66 1.32 Provides proprietary, SNMP and JMX based protocols for monitorability of the cluster. The proprietary protocols are supported via browser based and CLI interfaces 66 Reasonable Documentation 66 1.32 Good. There is a general resistence in Enterprises for MongoDB. 66 1.32 Excellent documentation. 99 Backup & Recovery are good. Even point in time recovery can be done 99 1.98 Providers a GUI to run the backup. MMS Backup Service. 99 1.98 Details are not investigated but can be recovered 66 It is relatively easier to engineer and deploy MarkLogic 33 0.66 Easy 33 0.66 Excellent documentation helps to engineer swiftly 99 Very High Price Point 33 0.66 Fair. 66 1.32 Oracle products are generally moderately priced if not expensive. 66 No information on the customer base 33 0.66 It is expanding its customer base. 31% of customers only reported no issues according to Gartner. 66 1.32 This is evolving in Oracle and no information on customer base. 33 Commercial Licensing (Restricted free version is also available) 66 1.32 AGPL(Drivers:Apache). Enterprise Licensing gets costlier for bigger enterprise. 66 1.32 AGPL 3 99 Evolving 66 1.32 Excellent Professional Support 99 1.98 Fair. 66
- 14. Gartner's report indicates the company is moving in multiple technology direction and may make the resources too thin. 33 0.66 Fast Evolving into Mature Model and depth in one single database solution. MongoDB is aggressively expanding the partnership. But MongoDB is not effectively putting barries to stop the competition. 99 1.98 Broader Market and Depth in Database Technology 99 61 64
- 15. WS Desc Score WS Desc Score WS 1.98 key-value datastore mostly used as in-memory DB and pub-sub mechanism. Extremely fast compared to others but limited by RAM and easiest to configure for small applications. No mobile support 99 1.98 Open-source, fault-tolerant key- value NoSQL database implementing principles from Amazon's Dynamo paper influenced by CAP Theorem. 99 1.98 1.32 For rapidly changing data with a foreseeable database size (should fit mostly in memory). OLTP and you can have a separate persistence DB or datawarehouse 66 1.32 Distributed database designed to deliver maximum data availability by distributing data across multiple servers across multiple data centers. High Resiliency due to server failure or network partition. 99 1.98 1.98 Stock prices. Analytics. Real-time data collection. Real-time communication. And wherever you used memcached before. 66 1.32 Content Management, Social applications, High Read/Write, simple applications. 66 1.32 1.32 Key-Value inmemory 99 1.98 Distributed Key-Value Store 66 1.32 RiakRedisSQL
- 16. 1.98 in-memort data structure store, Blazing fast 99 1.98 Own distributed full-text search engine with robust query language Fault-tolerant availability Queries Predictable latency Operational simplicity 99 1.98 1.32 Volatile memory, File System 99 1.98 Uses a simple key/value model for object storage. Objects in Riak consist of a unique key and a value, stored in a flat namespace called a bucket. You can store anything you want in Riak: text, images, JSON/XML/HTML documents, user and session data, backups, log files etc. 66 1.32 1.98 Yes 99 1.98 Yes 99 1.98 0.66 Possible to integrate with app coding(Need confirmation) 33 0.66 Native Search as well as Solr can be used 66 1.32 0.66 Yes 99 1.98 Utilizes LevelDB for compression. 99 1.98 1.98 Yes 99 1.98 Yes (Need to clarify) 99 1.98 1.98 Yes 99 1.98 Yes ( Need to clarify) 99 1.98 1.32 No 0 0 Yes (Supports for semantics in that MarkLogic can store RDF triples, using SPARQL as its query language.) 66 1.32
- 17. 1.32 API calls, Lua 66 1.32 Has official drivers for Ruby, Java, Erlang, Python, PHP, and C/C++ 99 1.98 1.32 C 66 1.32 Erlang, C, C++, some JavaScript, MapReduce 99 1.98 1.98 Yes can be used 66 1.32 JSON 66 1.32 1.32 Telnet-like, Binary safe 66 1.32 Utilizes PBC (Protocol Buffer Clients)interface, HTTP 66 1.32 1.32 No 0 0 Yes 66 1.32 1.98 Atomicity and consistency can be guaranteed for a group of commands with a server-side Lua script. Isolation is always guaranteed at command level, and can also be guaranteed for a group of command using a MULTI/EXEC block or a Lua script. Durability can be guaranteed when AOF is activated (with systematic fsync). Can be SPOF 66 1.32 CAP Theorem (Consistency, Availability, Partition tolerance (failure tolerance).) Riak focuses on Availability and Partition tolerance and falls more on the "eventually consistent" category. The theorem states only two out of the three properties can be fully relied on at any time. 66 1.32
- 18. 1.98 Atomicity and consistency can be guaranteed for a group of commands with a server-side Lua script. Isolation is always guaranteed at command level, and can also be guaranteed for a group of command using a MULTI/EXEC block or a Lua script. Durability can be guaranteed when AOF is activated (with systematic fsync) AI(C- Eventual Consistency - store to another DB, D- No, data is lost if hard disk crashes. Used to store specific time period data) 99 1.98 Does not support ACID transactions. ID (A - N, C - Eventually consistent) 66 1.32 1.98 Yes 99 1.98 No (As of Riak 1.4, counters were released to allow developers to build more complex functionality on top of data stored as keys and values.) 33 0.66 0.66 No 33 0.66 No 33 0.66 1.98 No 33 0.66 Yes 99 1.98 1.98 No 33 0.66 Yes 99 1.98 1.98 No 33 0.66 Yes 66 1.32 0 No 33 0.66 Yes 99 1.98 0.66In memory implementation can give high throughput66 1.32 Fair 66 1.32 0.66 In-Memory 99 1.98 Insall the memory Backend 66 1.32 0.66 It is doable 66 1.32 Possible 66 1.32
- 19. 1.32 Some overhead involved 33 0.66 Needed 33 0.66 1.32 Fair 33 0.66 Write latency is poor 33 0.66 1.32 Master-Slave replication, Automatic failover 66 1.32 Multi-Datacenter replication(Multi Master or Master Slave?) 66 1.32 1.98 No 33 0.66 Has a pluggable backend for its core shard-partitioned storage, with the default storage backend being Bitcask. Schemaless design allows more scalability ease. 66 1.32 1.98 Unix-like OS(*NIX), Mac Os X, Windows 99 1.98 Windows, Solaris, Linux, OS X, BSD 99 1.98
- 20. 1.32 Redis Admin UI 66 1.32 Many open source, self-hosted, and service-based solutions for aggregating and analyzing statistics and log data for the purposes of monitoring, alerting, and trend analysis on a Riak cluster. 33 0.66 1.98 Very Little 33 0.66 Good 33 0.66 1.32 Backup can be done by various ways 33 0.66 Backups could be inconsistent which will be corrected by read repair 33 0.66 1.98 Relatively involved process 33 0.66 Some efforts are involved 33 0.66 1.32 Low 33 0.66 Low 33 0.66 0.66 Moderate 33 0.66 30% Fortune 500 Companies uses it. They also develop and contibute drivers. 99 1.98 1.98 BSD-License 99 1.98 Apache Licensing 2.0 ( Open Source) 66 1.32 1.32 Need more details but appears to be pretty established 33 0.66 Reasoable 33 0.66
- 21. 1.98 Potentially competion from CouchBase, MongoDB, Oracle NoSQL 33 0.66 The scope is limited to No-SQL Key- value product only so the company's prospect in the broader DBMS market will be very limited. Oracel aggressive entry into this market could be challenging key- value space. 33 0.66 62 51 57