ACID vs BASE


In 1970, Jim Gray defined ACID properties as a reliable means of implementing a transaction system.

Atomic – A transaction should be indivisible and it must be all or nothing even when the database goes down due to system failures.

Consistency – A transaction should ensure that the data remains valid at all times even in case of application errors and none of the defined rules and constraints on the data must be compromised.

Isolation – A transaction must be isolated from other transactions and intermediate results of a transaction must not be seen by other transactions that are concurrently in execution.

Durability - A transaction must preserve the state of completed transaction at all times even in the event of system failures.

Even though ACID stood the test of time for centralized database system, and we see wide implementation of two-phase commit for distributed databases, scaling databases that are distributed and keeping them highly available has always been a challenge like for online stores such as Amazon, E Bay etc.

CAP

In 2000, Eric Brewer made a conjecture that it is impossible for a web service implementation to guarantee Consistency, Availability and Partition Tolerance and in 2002 Seth Gilbert and Nancy Lynch of MIT published a theorem based on Eric’s conjecture.

The alternative to ACID, to achieve high availability is called BASE (Basically available, Soft state, and eventually consistent.)

ACID is pessimistic and adheres to consistency after every operation.
BASE is optimistic and ensures the consistency towards the end.

ACID cancels any changes in the event of failure.
BASE does not cancel any changes when a partition fails.

ACID is difficult to scale and low availability compared to BASE
BASE is highly scalable and highly available

ACID implementation of a transaction is lot simpler.
BASE requires a more detailed analysis and scrutiny of the operations within a transaction.

In ACID, consistency as a property cannot be compromised
In BASE, consistency is compromised is for availability, however eventually the transaction are made consistent.

ACID, the state of any information is consistent and it is immediate right after the transaction is committed.

BASE, the state of any information can change over a period of time until it finally and
Eventually becomes consistent and hence the term Soft State, Eventually Consistent.

References.

http://www.sigmod.org/publications/sigmod-record/0806/p33.stonebraker.pdf


Comments

Post a Comment

Popular posts from this blog

Create SAS Data Sets w/o SAS

Image
Scope:  Create output in SAS7BDAT (closed binary format for passionate SAS practitioners and experts without installing Base SAS product from SAS Programming Institute. Carolina S-JDBC Driver: Dulles Research – A Virginia based private company offer “Carolina S-JDBC” JDBC driver that replace the need for Base SAS license. License constraints: (Carolina S-JDBC Driver Download, 2016) Annual subscription is $995.00 for a single instance of application server. Quote can be requested from enterprise wide installation. Free for non-commercial use. http://www.dullesresearch.com/sjdbc-download/ Solution:   class. forName ( "com.dullesopen.jdbc.Driver" ).newInstance();               Connection connection = DriverManager. getConnection (                       "jdbc:carolina:v2:libnames= ...
Read more

Immutable: String

The following are the some of the programmer friendly advantages of high level languages like Java, C# or Visual Basic by handling String as immutable as opposed to being mutable in languages like “C” A) Efficient use of memory: High level language make efficient use of memory by incorporating a special type of memory structure called "String Pool", each time any code creates a string literal, the Java Virtual Machine (Java) or Common language runtime checks to see if the String is already defined in the "StringPool" and if it exists, a reference is assigned as opposed to allocating space for every string. For example: If "ABCDEFGHIJKLMNOPQRSTUVWXYZ" is assigned10 times within a code, then in the StringPool, there will be a single copy of “ABCDEFGHIJKLMNOPQRSTUVWXYZ" and 10 references to it. This is a gain in the memory space allocation even though there is additional over head of maintaining “StringPool” memory structure. In ord...
Read more