Archive for the ‘SQL’ Category

The Twelve Days of NoSQL: Day Ten: Big Data in a Nutshell

January 4, 2014 2 comments

On the tenth day of Christmas, my true love gave to me
Ten lords a-leaping.

(Yesterday: NoSQL Taxonomy)(Tomorrow: Mistakes of the relational camp)

The topic of Big Data is often brought up in NoSQL discussions so let’s give it a nod. In 1998, Sergey Brin and Larry Page invented the PageRank algorithm for ranking web pages (The Anatomy of a Large-Scale Hypertextual Web Search Engine by Brin and Page) and founded Google. The PageRank algorithm required very large matrix-vector multiplications (Mining of Massive Datasets Ch. 5 by Rajaraman and Ullman) so the MapReduce technique was invented to handle such large computations (MapReduce: Simplified Data Processing on Large Clusters). Smart people then realized that the MapReduce technique could be used for other classes of problems and an open-source project called Hadoop was created to popularize the MapReduce technique (The history of Hadoop: From 4 nodes to the future of data). Other smart people realized that MapReduce could handle the operations of relational algebra such as join, anti-join, semi-join, union, difference, and intersection (Mining of Massive Datasets Ch. 2 by Rajaraman and Ullman) and began looking at the possibility of processing large volumes of business data (a.k.a. “Big Data”) better and cheaper than mainstream database management systems. Initially programmers had to write Java code for the “mappers” and “reducers” used by MapReduce. However, smart people soon realized that SQL queries could be automatically translated into the necessary Java code and “SQL-on-Hadoop” was born. Big Data thus became about processing large volumes of business data with SQL but better and cheaper than mainstream database management systems. However, the smart people have now realized that MapReduce is not the best solution for low-latency queries (Facebook open sources its SQL-on-Hadoop engine, and the web rejoices). Big Data has finally become about processing large volumes of business data with SQL but better and cheaper than mainstream database management systems and with or without MapReduce.

That’s the fast-moving story of Big Data in a nutshell.

Also see: The Twelve Days of SQL: Day Ten: Sometimes the optimizer needs a hint

The Twelve Days of NoSQL: Day Nine: NoSQL Taxonomy

On the ninth day of Christmas, my true love gave to me
Nine ladies dancing.

(Yesterday: Oracle NoSQL Database)(Tomorrow: Big Data in a Nutshell)

NoSQL databases can be classified into the following categories:

  • Key-value stores: The archetype is Amazon Dynamo of which DynamoDB is the commercial successor. Key-value stores basically allow applications to “put” and “get” values but each product has differentiators. For example, DynamoDB supports “tables” (namespaces) while Oracle NoSQL Database offers “major” and “minor” key paths.
  • Document stores: While key-value stores treat values as uninterpreted strings, document stores allow values to be managed using formats such as JSON (JavaScript Object Notation) which are conceptually similar to XML. This allows key-value pairs to be indexed by any component of the value just as XML data can be indexed in mainstream database management systems.
  • Column-family stores: Column-family stores allow data associated with a single key to be spread over multiple storage nodes. Each storage node only stores a subset of the data associated with the key; hence the name “column-family.” A key is therefore composed of a “row key” and a “column key” in a manner analogous to the major and minor key paths of Oracle NoSQL Database.
  • Graph databases: Graph databases are non-relational databases that use graph concepts such as nodes and edges to solve certain classes of problems: for example; the shortest route between two towns on a map. The concepts of functional segmentation, sharding, replication, eventual consistency, and schemaless design do not apply to graph databases so I will not discuss graph databases.

NoSQL products are numerous and rapidly evolving. There is a crying need for a continuously updated encyclopedia of NoSQL products but none exists. There is a crying need for an independent benchmarking organization but none exists. My best advice is to do a proof of concept (POC) as well as a PSR (Performance Scalability Reliability) test before committing to using a NoSQL product. Back in the day, in 1985 to be precise, Dr. Codd had words of advice for those who were debating between the new relational products and the established pre-relational products of his day. The advice is as solid today as it was in Dr. Codd’s day.

“Any buyer confronted with the decision of which DBMS to acquire should weigh three factors heavily.

The first factor is the buyer’s performance requirements, often expressed in terms of the number of transactions that must be executed per second. The average complexity of each transaction is also an important consideration. Only if the performance requirements are extremely severe should buyers rule out present relational DBMS products on this basis. Even then buyers should design performance tests of their own, rather than rely on vendor-designed tests or vendor-declared strategies. [emphasis added]

The second factor is reduced costs for developing new databases and new application programs …

The third factor is protecting future investments in application programs by acquiring a DBMS with a solid theoretical foundation …

In every case, a relational DBMS wins on factors two and three. In many cases, it can win on factor one also—in spite of all the myths about performance.”

—An Evaluation Scheme for Database Management Systems that are claimed to be Relational

Also see: The Twelve Days of SQL: Day Nine: Physical database design matters

Categories: DBA, NoSQL, Oracle, SQL, Uncategorized Tags: , , ,

The Twelve Days of NoSQL: Day Seven: Schemaless Design

December 31, 2013 Leave a comment

On the seventh day of Christmas, my true love gave to me
Seven swans a-swimming.

(Yesterday: The False Premise of NoSQL)(Tomorrow: Oracle NoSQL Database)

As we discussed on Day One, NoSQL consists of “disruptive innovations” that are gaining steam and moving upmarket. So far, we have discussed functional segmentation (the pivotal innovation), sharding, asynchronous replication, eventual consistency (resulting from lack of distributed transactions across functional segments and from asynchronous replication), and blobs.

The final innovation of the NoSQL camp is “schemaless design.” In database management systems of the NoSQL kind, data is stored in “blobs” and documents the database management system does not police their structure. In mainstream database management systems on the other hand, doctrinal purity requires that the schema be designed before data is inserted. Let’s do a thought experiment.

Suppose that we don’t have a schema and let’s suppose that the following facts are known.

  • Iggy Fernandez is an employee with EMPLOYEE_ID=1 and SALARY=$1000.
  • Mogens Norgaard is a commissioned employee with EMPLOYEE_ID=2, SALARY=€1000, and COMMISSION_PCT=25.
  • Morten Egan is a commissioned employee with EMPLOYEE_ID=3, SALARY=€1000, and unknown COMMISSION_PCT.

Could we ask the following questions and expect to receive correct answers?

  • Question: What is the salary of Iggy Fernandez?
  • Correct answer: $1000.
  • Question: What is the commission percentage of Iggy Fernandez?
  • Correct answer: Invalid question.
  • Question: What is the commission percentage of Mogens Norgaard?
  • Correct answer: 25%
  • Question: What is the commission percentage of Morten Egan?
  • Correct answer: Unknown.

If we humans can process the above data and correctly answer the above questions, then surely we can program computers to do so.

The above data could be modeled with the following three relations. It is certainly disruptive to suggest that this be done on the fly by the database management system but not outside the realm of possibility.




A NoSQL company called Hadapt has already stepped forward with such a feature:

“While it is true that SQL requires a schema, it is entirely untrue that the user has to define this schema in advance before query processing. There are many data sets out there, including JSON, XML, and generic key-value data sets that are self-describing — each value is associated with some key that describes what entity attribute this value is associated with [emphasis added]. If these data sets are stored in Hadoop, there is no reason why Hadoop cannot automatically generate a virtual schema against which SQL queries can be issued. And if this is true, users should not be forced to define a schema before using a SQL-on-Hadoop solution — they should be able to effortlessly issue SQL against a schema that was automatically generated for them when data was loaded into Hadoop.

Thanks to the hard work of many people at Hadapt from several different groups, including the science team who developed an initial design of the feature, the engineering team who continued to refine the design and integrate it into Hadapt’s SQL-on-Hadoop solution, and the customer solutions team who worked with early customers to test and collect feedback on the functionality of this feature, this feature is now available in Hadapt.” (

This is not really new ground. Oracle Database provides the ability to convert XML documents into relational tables ( though it ought to be possible to view XML data as tables while physically storing it in XML format in order to benefit certain use cases. It should also be possible to redundantly store data in both XML and relational formats in order to benefit other use cases.

In  “Extending the Database Relational Model to Capture More Meaning,” Dr. Codd explains how a “formatted database” is created from a collection of facts:

“Suppose we think of a database initially as a set of formulas in first-order predicate logic. Further, each formula has no free variables and is in as atomic a form as possible (e.g, A & B would be replaced by the component formulas A, B). Now suppose that most of the formulas are simple assertions of the form Pab…z (where P is a predicate and a, b, … , z are constants), and that the number of distinct predicates in the database is few compared with the number of simple assertions. Such a database is usually called formatted, because the major part of it lends itself to rather regular structuring. One obvious way is to factor out the predicate common to a set of simple assertions and then treat the set as an instance of an n-ary relation and the predicate as the name of the relation.”

In other words, a collection of facts can always be organized into relations if necessary.

Also see: The Twelve Days of SQL: Day Seven: EXPLAIN PLAN lies

The Twelve Days of NoSQL: Day Six: The False Premise of NoSQL

December 30, 2013 5 comments

On the sixth day of Christmas, my true love gave to me
Six geese a-laying.

(Yesterday: Replication and Eventual Consistency)(Tomorrow: Schemaless Design)

The final hurdle was extreme performance, and that’s where the Dynamo developers went astray. The Dynamo developers believed that the relational model imposes a “join penalty” and therefore chose to store data as “blobs.” This objection to the relational model is colorfully summarized by the following statement attributed to Esther Dyson, the editor of the Release 1.0 newsletter, “Using tables to store objects is like driving your car home and then disassembling it to put it in the garage. It can be assembled again in the morning, but one eventually asks whether this is the most efficient way to park a car.” [1] The statement dates back to 1988 and was much quoted when object-oriented databases were in vogue.

Since the shopping cart is an object, doesn’t disassembling it for storage make subsequent data retrieval and updates inefficient? The belief stems from an unfounded assumption that has found its way into every mainstream DBMS—that every table should map to physical storage. In reality, the relational model is a logical model and, therefore, it does not concern itself with storage details at all. It would be perfectly legitimate to store the shopping cart in a physical form that resembled a shopping cart while still offering a relational model of the data complete with SQL. In other words, the physical representation could be optimized for the most important use case—retrieving the entire shopping-cart object using its key—without affecting the relational model of the data. It would also be perfectly legitimate to provide a non-relational API for the important use cases. Dr. Codd himself gave conditional blessing to such non-relational APIs in his 1985 Computerworld article, “Is Your DBMS Really Relational?”, in which he says, If a relational system has a low-level (single-record-at-a-time) language, that low level [should not] be used to subvert or bypass the integrity rules and constraints expressed in the higher level relational language (multiple-records-at-a-time).”

The key-blob or “key-value” approach used by Dynamo and successor products would be called “zeroth” normal form in relational terminology. In his 1970 paper, Dr. Codd says: Nonatomic values can be discussed within the relational framework. Thus, some domains may have relations as elements. These relations may, in turn, be defined on nonsimple domains, and so on. For example, one of the domains on which the relation employee is defined might be salary history. An element of the salary history domain is a binary relation defined on the domain date and the domain salary. The salary history domain is the set of all such binary relations. At any instant of time there are as many instances of the salary history relation in the data bank as there are employees. In contrast, there is only one instance of the employee relation.” In common parlance, a relation with non-simple domains is said to be in “zeroth” normal form or unnormalized. Dr. Codd suggested that unnormalized relations should be normalized for ease of use. Here again is the unnormalized employee relation from Dr. Codd’s paper:

employee (
  jobhistory (jobdate, title, salaryhistory (salarydate, salary)),
  children (childname, birthyear)

The above unnormalized relation can be decomposed into four normalized relations as follows.

employee' (employee#, name, birthdate)
jobhistory' (employee#, jobdate, title)
salaryhistory' (employee#, jobdate, salarydate, salary)
children' (employee#, childname, birthyear)

However, this is not to suggest that these normalized relations must necessarily be mapped to individual buckets of physical storage. Dr. Codd differentiated between the stored set, the named set, and the expressible set. In the above example, we have one unnormalized relation and four normalized relations, if we preferred it, the unnormalized employee relation could be the only member of the stored set. Alternatively, if we preferred it, all five relations could be part of the stored set; that is, we could legitimately store redundant representations of the data. However, the common belief blessed by current practice is that the normalized relations should be the only members of the stored set.

Even if the stored set contains only normalized relations, they need not map to different buckets of physical storage. Oracle is unique among mainstream database management systems in providing a convenient construct called the “table cluster” that is suitable for hierarchical schemas. In Dr. Codd’s example, employee# would be the cluster key, and rows corresponding to the same cluster key from all four tables could be stored in the same physical block on disk thus avoiding the join penalty. If the cluster was a “hash cluster,” no indexes would be required for the use case of retrieving records belonging to a single cluster key.

The mistake made by the Dynamo developers is really a mistake perpetuated by the relational camp but it is a mistake nevertheless.

Tomorrow: Schemaless Design

1. I’ve been unable to find the statement in the Release 1.0 archives at so I don’t really know the true source or author of the statement. However, the statement is popularly attributed to Esther Dyson and claimed to have been published in the Release 1.0 newsletter. I found a claim that the statement is found in the September 1988 issue but that didn’t pan out.

Appendix: Table Clusters in Oracle Database

Here’s a demonstration of using Oracle table clusters to store records from four tables in the same block and retrieving all the components of the “employee cart” without using indexes. First we create four normalized tables and prove that all the records of a single employee including job history, salary history, and children are stored in a single database block so that there is never any join-penalty when assembling employee data. Then we create an object-relational view that assembles employee information into a single unnormalized structure and show how to insert into this view using an “INSTEAD OF” trigger.

The following demonstration was performed using a pre-Built developer VM for Oracle VM VirtualBox. The version of Oracle Database is

SQL*Plus: Release Production on Sun Jul 28 19:44:23 2013

Copyright (c) 1982, 2010, Oracle.  All rights reserved.

Connected to:
Oracle Database 11g Enterprise Edition Release - Production
With the Partitioning, OLAP, Data Mining and Real Application Testing options

First, we create a table cluster and add four tables to the cluster.

SQL> CREATE CLUSTER employees (employee# INTEGER) hashkeys 1000;

Cluster created.

  2  (
  3    employee# INTEGER NOT NULL,
  4    name VARCHAR2(16),
  5    birth_date DATE,
  6    CONSTRAINT employees_pk PRIMARY KEY (employee#)
  7  )
  8  CLUSTER employees (employee#);

Table created.

SQL> CREATE TABLE job_history
  2  (
  3    employee# INTEGER NOT NULL,
  4    job_date DATE NOT NULL,
  5    title VARCHAR2(16),
  6    CONSTRAINT job_history_pk PRIMARY KEY (employee#, job_date),
  7    CONSTRAINT job_history_fk1 FOREIGN KEY (employee#) REFERENCES employees
  8  )
  9  CLUSTER employees (employee#);

Table created.

SQL> CREATE TABLE salary_history
  2  (
  3    employee# INTEGER NOT NULL,
  4    job_date DATE NOT NULL,
  5    salary_date DATE NOT NULL,
  6    salary NUMBER,
  7    CONSTRAINT salary_history_pk PRIMARY KEY (employee#, job_date, salary_date),
  8    CONSTRAINT salary_history_fk1 FOREIGN KEY (employee#) REFERENCES employees,
  9    CONSTRAINT salary_history_fk2 FOREIGN KEY (employee#, job_date) REFERENCES job_history
 10  )
 11  CLUSTER employees (employee#);

Table created.

  2  (
  3    employee# INTEGER NOT NULL,
  4    child_name VARCHAR2(16) NOT NULL,
  5    birth_date DATE,
  6    CONSTRAINT children_pk PRIMARY KEY (employee#, child_name),
  7    CONSTRAINT children_fk1 FOREIGN KEY (employee#) REFERENCES employees
  8  )
  9  CLUSTER employees (employee#);

Table created.

Then we insert data into all four tables. We find that all the records have been stored in the same database block even though they belong to different tables. Therefore the join-penalty has been eliminated.

SQL> INSERT INTO employees VALUES (1, 'IGNATIUS', '01-JAN-1970');

1 row created.

SQL> INSERT INTO children VALUES (1, 'INIGA', '01-JAN-2001');

1 row created.

SQL> INSERT INTO children VALUES (1, 'INIGO', '01-JAN-2002');

1 row created.

SQL> INSERT INTO job_history VALUES (1, '01-JAN-1991', 'PROGRAMMER');

1 row created.

SQL> INSERT INTO job_history VALUES (1, '01-JAN-1992', 'DATABASE ADMIN');

1 row created.

SQL> INSERT INTO salary_history VALUES (1, '01-JAN-1991', '1-FEB-1991', 1000);

1 row created.

SQL> INSERT INTO salary_history VALUES (1, '01-JAN-1991', '1-MAR-1991', 1000);

1 row created.

SQL> INSERT INTO salary_history VALUES (1, '01-JAN-1992', '1-FEB-1992', 2000);

1 row created.

SQL> INSERT INTO salary_history VALUES (1, '01-JAN-1992', '1-MAR-1992', 2000);

1 row created.

SQL> SELECT DISTINCT DBMS_ROWID.ROWID_BLOCK_NUMBER(rowid) AS block_number FROM employees where employee# = 1;


SQL> SELECT DISTINCT DBMS_ROWID.ROWID_BLOCK_NUMBER(rowid) AS block_number FROM children where employee# = 1;


SQL> SELECT DISTINCT DBMS_ROWID.ROWID_BLOCK_NUMBER(rowid) AS block_number FROM job_history where employee# = 1;


SQL> SELECT DISTINCT DBMS_ROWID.ROWID_BLOCK_NUMBER(rowid) AS block_number FROM salary_history where employee# = 1;


Next we create an object-relational view that presents each employee as an object.

SQL> CREATE OR REPLACE TYPE children_rec AS OBJECT (child_name VARCHAR2(16), birth_date DATE)
  2  /

Type created.

SQL> CREATE OR REPLACE TYPE children_tab AS TABLE OF children_rec
  2  /

Type created.

SQL> CREATE OR REPLACE TYPE salary_history_rec AS OBJECT (salary_date DATE, salary NUMBER)
  2  /

Type created.

SQL> CREATE OR REPLACE TYPE salary_history_tab AS TABLE OF salary_history_rec
  2  /

Type created.

SQL> CREATE OR REPLACE TYPE job_history_rec AS OBJECT (job_date DATE, title VARCHAR2(16), salary_history SALARY_HISTORY_TAB)
  2  /

Type created.

SQL> CREATE OR REPLACE TYPE job_history_tab AS TABLE of job_history_rec
  2  /

Type created.

SQL> create or replace view employees_view as
  3    employee#,
  4    name,
  5    birth_date,
  6    CAST
  7    (
  8      MULTISET
  9      (
 10        SELECT
 11          child_name,
 12          birth_date
 13        FROM children
 14        WHERE employee#=e.employee#
 15      )
 16      AS children_tab
 17    ) children,
 18    CAST
 19    (
 20      MULTISET
 21      (
 22        SELECT
 23          job_date,
 24          title,
 25          CAST
 26          (
 27            MULTISET
 28            (
 29              SELECT salary_date, salary
 30              FROM salary_history
 31              WHERE employee#=e.employee#
 32              AND job_date=jh.job_date
 33            )
 34            AS salary_history_tab
 35          ) salary_history
 36        FROM job_history jh
 37        WHERE employee#=e.employee#
 38      )
 39      AS job_history_tab
 40    ) job_history
 41  FROM employees e;

View created.

Let’s retrieve one employee object and look at the query execution plan. No indexes are used in retrieving records from each of the four tables. The cost of the plan is just 1. This is the minimum achievable cost, indicating that there is no join-penalty.

SQL> alter session set "_rowsource_execution_statistics"=true;

Session altered.

SQL> SELECT * FROM employees_view WHERE employee# = 1;

---------- ---------------- ---------
         1 IGNATIUS         01-JAN-70
ALARY_HISTORY_REC('01-MAR-92', 2000))))

SQL> SELECT * FROM TABLE(dbms_xplan.display_cursor(null, null, 'TYPICAL IOSTATS LAST'));

SQL_ID  aaxmaqz947aa0, child number 0
SELECT * FROM employees_view WHERE employee# = 1

Plan hash value: 2117652374

| Id  | Operation         | Name           | Starts | E-Rows |E-Bytes| Cost  | A-Rows |   A-Time   | Buffers |
|   0 | SELECT STATEMENT  |                |      1 |        |       |     1 |      1 |00:00:00.01 |       1 |
|*  1 |  TABLE ACCESS HASH| CHILDREN       |      1 |      1 |    32 |       |      2 |00:00:00.01 |       1 |
|*  2 |  TABLE ACCESS HASH| SALARY_HISTORY |      2 |      1 |    44 |       |      4 |00:00:00.01 |       3 |
|*  3 |  TABLE ACCESS HASH| JOB_HISTORY    |      1 |      1 |    32 |       |      2 |00:00:00.01 |       1 |
|*  4 |  TABLE ACCESS HASH| EMPLOYEES      |      1 |    845 | 27040 |       |      1 |00:00:00.01 |       1 |

Predicate Information (identified by operation id):

   1 - access("EMPLOYEE#"=:B1)
   2 - access("EMPLOYEE#"=:B1)
   3 - access("EMPLOYEE#"=:B1)
   4 - access("EMPLOYEE#"=1)

   - cpu costing is off (consider enabling it)
   - dynamic sampling used for this statement (level=2)

30 rows selected.

Next, let’s create an “INSTEAD OF” trigger so that we insert into the view directly; that is, use a single insert statement instead of multiple insert statements. The trigger will do all the heavy-lifting for us.

SQL> CREATE OR REPLACE TRIGGER employees_view_insert
  2  INSTEAD OF INSERT ON employees_view
  6    i NUMBER;
  7  BEGIN
  8    INSERT INTO employees
  9    VALUES
 10    (
 11      :n.employee#,
 13      :n.birth_date
 14    );
 16    FOR i IN :n.children.FIRST .. :n.children.LAST
 17    LOOP
 18      INSERT INTO children
 19      VALUES
 20      (
 21        :n.employee#,
 22        :n.children(i).child_name,
 23        :n.children(i).birth_date
 24      );
 25    END LOOP;
 27    FOR i IN :n.job_history.FIRST .. :n.job_history.LAST
 28    LOOP
 29      INSERT INTO job_history VALUES
 30      (
 31        :n.employee#,
 32        :n.job_history(i).job_date,
 33        :n.job_history(i).title
 34      );
 35      FOR j IN :n.job_history(i).salary_history.FIRST .. :n.job_history(i).salary_history.LAST
 36      LOOP
 37        INSERT INTO salary_history
 38        VALUES
 39        (
 40          :n.employee#,
 41          :n.job_history(i).job_date,
 42          :n.job_history(i).salary_history(j).salary_date,
 43          :n.job_history(i).salary_history(j).salary
 44         );
 45      END LOOP;
 46    END LOOP;
 47  END;
 48  /

Trigger created.

Finally, let’s insert an employee object directly into the view and confirm that we can read it back.

SQL> INSERT INTO employees_view
  3  (
  4    2,
  5    'YGNACIO',
  6    '01-JAN-70',
  8    (
  9      CHILDREN_REC('INIGA', '01-JAN-01'),
 10      CHILDREN_REC('INIGO', '01-JAN-02')
 11    ),
 13    (
 15      (
 16        '01-JAN-91',
 17        'PROGRAMMER',
 19        (
 20          SALARY_HISTORY_REC('01-FEB-91', 1000),
 21          SALARY_HISTORY_REC('01-MAR-91', 1000)
 22        )
 23      ),
 25      (
 26        '01-JAN-92',
 27        'DATABASE ADMIN',
 29        (
 30          SALARY_HISTORY_REC('01-FEB-92', 2000),
 31          SALARY_HISTORY_REC('01-MAR-92', 2000)
 32        )
 33      )
 34    )
 35  );

1 row created.

SQL> SELECT * FROM employees_view WHERE employee# = 2;

---------- ---------------- ---------
         2 YGNACIO          01-JAN-70
ALARY_HISTORY_REC('01-MAR-92', 2000))))

Also see: The Twelve Days of SQL: Day Six: The execution plan is a tree

The Twelve Days of NoSQL: Day Five: Replication and Eventual Consistency

December 29, 2013 Leave a comment

On the fifth day of Christmas, my true love gave to me
Five golden rings.

(Yesterday: Sharding)(Tomorrow: The False Premise of NoSQL)

By now, you must be wondering when I’m going to get around to explaining how to create a NoSQL database. When I was a junior programmer, quite early in my career, my friends and I were assigned to work on a big software development project for which we would have to use technologies with which we were completely unfamiliar. We were promised that training would be provided before the project started. The very first thing the instructor said was (paraphrasing) “First you have to insert your definitions into the C.D.D.” and he walked to the board and wrote the commands that we needed for the purpose. Needless to say, we were quite flustered because we had no idea what those “definitions” might be or what a “C.D.D.” was and how it fit into the big picture.

NoSQL is being taught without reference to the big picture. None of the current books on NoSQL mention functional segmentation even though it is the underpinning principle of NoSQL. All the current books on NoSQL imply that NoSQL principles are in conflict with the relational model. If you are in a hurry to create your first NoSQL database, I can recommend Seven Databases in Seven Weeks: A Guide to Modern Databases and the NoSQL Movement. But as one of the world’s greatest geniuses Leonardo da Vinci has said: Those who are in love with practice without science are like the sailor who gets into a ship without rudder or compass, who is never certain where he is going. Practice must always be built on sound theory … The painter who copies by practice and judgement of eye, without rules, is like a mirror which imitates within itself all the things placed before it without any understanding of them.” (On the errors of those who rely on practice without science).

Continuing the train of thought from Day Four, the Dynamo developers saw that one of the keys to extreme availability was data replication. Multiple copies of the shopping cart are allowed to exist and, if one of the replicas becomes unresponsive, the data can be served by one of the other replicas. However, because of network latencies, the copies may occasionally get out of sync and the customer may occasionally encounter a stale version of the shopping cart. Once again, this can be handled appropriately by the application tier; the node that falls behind can catch up eventually or inconsistencies can be detected and resolved at an opportune time, such as at checkout. This technique is called “eventual consistency.”

The inventor of relational theory, Dr. Codd, was acutely aware of the potential overhead of consistency checking. In his 1970 paper, he said:

“There are, of course, several possible ways in which a system can detect inconsistencies and respond to them. In one approach the system checks for possible inconsistency whenever an insertion, deletion, or key update occurs. Naturally, such checking will slow these operations down. [emphasis added] If an inconsistency has been generated, details are logged internally, and if it is not remedied within some reasonable time interval, either the user or someone responsible for the security and integrity of the data is notified. Another approach is to conduct consistency checking as a batch operation once a day or less frequently.”

In other words, the inventor of relational theory would not have found a conflict between his relational model and the “eventual consistency” that is one of the hallmarks of the NoSQL products of today. However, the Dynamo developers imagined a conflict because it quite understandably conflated the relational model with the ACID guarantees of database management systems. However, ACID has nothing to do with the relational model per se (although relational theory does come in very handy in defining consistency constraints); pre-relational database management systems such as IMS provided ACID guarantees and so did post-relational object-oriented database management systems.

I should not defend eventual consistency simply by using a convenient quote from the writings of Dr. Codd. “The devil can cite Scripture for his purpose. An evil soul producing holy witness is like a villain with a smiling cheek, a goodly apple rotten at the heart. O, what a goodly outside falsehood hath!” (from the Shakespeare play The Merchant of Venice) If I am in favor of eventual consistency, I should explain why, not simply quote from the writings of Dr. Codd. If I can defend my own beliefs, I free myself to disagree with Dr. Codd as I plan to do later in this series. I have in fact come to accept that real-time consistency checking should be a design choice not a scriptural mandate. I may have had a different opinion in the past but a foolish consistency is the hobgoblin of little minds, adored by little statesmen and philosophers and divines. … Speak what you think now in hard words, and to-morrow speak what to-morrow thinks in hard words again, though it contradict every thing you said to-day.” (from the Emerson essay Self-Reliance).

The tradeoff between consistency and performance is as important in the wired world of today as it was in Dr. Codd’s world. We cannot cast stones at Dynamo for the infraction of not guaranteeing the synchronization of replicated data (or allowing temporary inconsistencies between functional segments), because violations of the consistency requirement are equally commonplace in the relational camp. The replication technique used by Dynamo has a close parallel in the technique of “multimaster replication” used in the relational camp. Application developers in the relational camp are warned about the negative impact of integrity constraints. [1] [2] [3] [4] And, most importantly, no mainstream DBMS currently implements the SQL-92 “CREATE ASSERTION” feature that is necessary to provide the consistency guarantee. For a detailed analysis of this anomaly, refer to Toon Koppelaars’s article “CREATE ASSERTION: The Impossible Dream” in the August 2013 issue of the NoCOUG Journal.

Also see: The Twelve Days of SQL: Day Five: The query cost is only an estimate

1. “Using primary and foreign keys can impact performance. Avoid using them when possible.” (

2. “For performance reasons, the Oracle BPEL Process Manager, Oracle Mediator, human workflow, Oracle B2B, SOA Infrastructure, and Oracle BPM Suite schemas have no foreign key constraints to enforce integrity.” (

3. “For database independence, applications typically do not store the primary key-foreign key relationships in the database itself; rather, the relationships are enforced in the application.” (

4. “The ETL process commonly verifies that certain constraints are true. For example, it can validate all of the foreign keys in the data coming into the fact table. This means that you can trust it to provide clean data, instead of implementing constraints in the data warehouse.” (

The Twelve Days of NoSQL: Day Four: Sharding

December 28, 2013 Leave a comment

On the fourth day of Christmas, my true love gave to me
Four colly birds.

(Yesterday: Functional Segmentation)(Tomorrow: Replication and Eventual Consistency)

Let’s recap what we have covered so far.

  • Day One: NoSQL technology consists of “disruptive innovations” which present a dilemma for established players.
  • Day Two: The goals of NoSQL technology are extreme performance, extreme scalability, and extreme availability.
  • Day Three: The underpinning component of NoSQL technology is functional segmentation which results in simple hierarchical schemas.

Amazon’s next design decision was “sharding” or horizontal partitioning of all the tables in a hierarchical schema. Hash-partitioning is typically used. Each table is partitioned in the same way as the other tables in the schema and each set of partitions is placed in a separate database referred to as a “shard.” The shards are independent of each other; that is, there is no clustering (as in Oracle RAC) or federation (as in IBM DB2).

Note that the hierarchical schemas that result from functional segmentation are always shardable; that is, hierarchical schemas are shardable by definition.

Returning to the example from Ted Codd’s 1970 paper on the relational model:

  • employee (man#, name, birthdate) with primary key (man#)
  • children (man#, childname, birthyear) with primary key (man#, childname)
  • jobhistory (man#, jobdate, title) with primary key (man#, jobdate)
  • salaryhistory (man#, jobdate, salarydate, salary) with primary key (man#, jobdate, salarydate)

Note that the jobhistory, salaryhistory, and children tables have composite keys. In each case, the leading column of the composite key is the man#. Therefore, all four tables can be partitioned using the man#.

Sharding is an essential component of NoSQL designs but it does not present a conflict with the relational model; it too is simply a physical database design decision. In the relational model, the collection of standalone databases or shards can be logically viewed as a single distributed database.

Also see: The Twelve Days of SQL: Day Four: The way you write your query matters

The Twelve Days of NoSQL: Day Three: Functional Segmentation

December 27, 2013 4 comments

On the third day of Christmas, my true love gave to me
Three French hens.

(Yesterday: Requirements and Assumptions)(Tomorrow: Sharding)

Amazon’s pivotal design decision was to break its monolithic enterprise-wide database service into simpler component services such as a best-seller list service, a shopping cart service, a customer preferences service, a sales rank service, and a product catalog service. This avoided a single point of failure. In an interview for the NoCOUG Journal, Amazon’s first database administrator, Jeremiah Wilton explains the rationale behind Amazon’s approach: The best availability in the industry comes from application software that is predicated upon a surprising assumption: The databases upon which the software relies will inevitably fail. The better the software’s ability to continue operating in such a situation, the higher the overall service’s availability will be. But isn’t Oracle unbreakable? At the database level, regardless of the measures taken to improve availability, outages will occur from time to time. An outage may be from a required upgrade or a bug. Knowing this, if you engineer application software to handle this eventuality, then a database outage will have less or no impact on end users. In summary, there are many ways to improve a single database’s availability. But the highest availability comes from thoughtful engineering of the entire application architecture.” As an example, the shopping cart service should not be affected if the checkout service is unavailable or not performing well.

I said that this was the pivotal design decision made by Amazon. I cannot emphasize this enough. If you resist functional segmentation, you are not ready for NoSQL. If you miss the point, you will not understand NoSQL.

Note that functional segmentation results in simple hierarchical schemas. Here is an example of a simple hierarchical schema from Ted Codd’s 1970 paper on the relational model, meticulously reproduced in the 100th issue of the NoCOUG Journal. This schema stores information about employees, their children, their job histories, and their salary histories.

  • employee (man#, name, birthdate)
  • children (man#, childname, birthyear)
  • jobhistory (man#, jobdate, title)
  • salaryhistory (man#, jobdate, salarydate, salary)

Functional segmentation is the underpinning of NoSQL technology but it does not present a conflict with the relational model; it is simply a physical database design decision. Each functional segment is usually assigned its own standalone database. The collection of functional segments could be regarded as a single distributed database. However, distributed transactions are verboten in the NoSQL world. Functional segmentation can therefore result in temporary inconsistencies if, for example, the shopping cart data is not in the same database as the product catalog and occasional inconsistencies result. Occasionally, an item that is present in a shopping cart may go out of stock. Occasionally, an item that is present in a shopping cart may be repriced. The problems can be resolved when the customer decides to check out, if not earlier. As an Amazon customer, I occasionally leave items in my shopping cart but don’t complete a purchase. When I resume shopping, I sometimes get a notification that an item in my shopping chart is no longer in stock or has been repriced. This technique is called “eventual consistency” and the application is responsible for ensuring that inconsistencies are eventually corrected. Randy Shoup, one of the architects of the eBay’s ecommerce platform, explains how:

“At eBay, we allow absolutely no client-side or distributed transactions of any kind – no two-phase commit. In certain well-defined situations, we will combine multiple statements on a single database into a single transactional operation. For the most part, however, individual statements are auto-committed. While this intentional relaxation of orthodox ACID properties does not guarantee immediate consistency everywhere, the reality is that most systems are available the vast majority of the time. Of course, we do employ various techniques to help the system reach eventual consistency: careful ordering of database operations, asynchronous recovery events, and reconciliation or settlement batches. We choose the technique according to the consistency demands of the particular use case.” (Scalability Best Practices: Lessons from eBay)

The eventual consistency technique receives a lot of attention because it is supposedly in conflict with the relational model. We will return to this subject later in this series and argue that eventual consistency is not in conflict with the relational model.

Also see: The Twelve Days of SQL: Day Three: There isn’t always a single optimal query plan for a SQL query

%d bloggers like this: