Distributed Databases

What is a distributed database?

• A standard database contains a central database that can be accessed from a workstation over a network. All traffic has to flow over the network which can lead to problems.
• A typical distributed database is split across different sites and each site is connected over a network.
• Queries and updates can be made locally on the site through a database management system (DBMS). Each site also has a distributed database component (DDBMS) that allows each site to make global queries of the whole system.
• 
• Bildunterschrift: Standard Database and Distributed Database

Why use a distributed database?

• Many organisations have branches and offices located across the country or even the world.
• For example:
• A travel company with many branches across the country
• An international car maker with with dealers across the world
• A supermarket stock system with many branches
• Branches may only need to access their own data 90% of the time and may only occasionally need access to remote information. The central office will need to have an up to date version of all the data as well.
• Queries and updates in a single central database will have to travel a network. Heavy load can be put on the system due to this. It also means if the central database breaks or the network goes down, the entire organisation is affected.
• A distributed database solves some of these issues.
• The central database is updated over night as a batch operation, but day-to-day work in the branches is carried out using the local portion of the database. If the network fails, then they can carry on as normal until it is fixed.

Duplicated database at each location

• With this system, a local copy of the entire database is kept at each location.
• This is often not practical for a number of reasons:
• Disadvantages:
• Very heavy use of the network
• Whole database may be huge requiring large storage requirement at each location
• Difficult to keep all copies in synchronisation
• It does not scale well as the database grows
• Wasteful if only a small area of the database is needed by each branch
• Advantages:
• A practical solution if the database is fairly small
• 
• Bildunterschrift: Duplicated database at each station

Partitioned database

• The distributed database is partitioned into convenient data sets depending on the specific needs of the organisation. A store may have its own section of the distributed stock control database. Each night, a central database is synchronised with records form each store.
• Advantages:
• Each store is independent. Problems in one store does not affect any other store.
• Network load is much less, as only the central database needs to be kept in sync.
• High performance as there is no network bottleneck.
• Very good solution if each store is largely independent.
• Scales up very easily.
• Disadvantages:
• Not so useful if each branch constantly needs information from other branches.
• The database must be carefully partitioned to keep each section as local as possible.
• 
• Bildunterschrift: Partitioned Database

Partitioned and Index

• 
• Bildunterschrift: Partitioned and Index
• A modification of the that partitions that database to keep all records as local as possible, but also contains indexes that point to records in other databases. A central database keeps the indexes updated on a nightly batch run.
• Advantages:
• Still high performance as most queries and updates remain local
• Can efficiently access remote records by using the index rather than a network query to a central database.
• Disadvantages:
• The system must now keep all indexes up to date
• More complicated

Security

• Security is more difficult to maintain compared to a central database.
• Multiple Entry Points
• One of the weaknesses of a distributed database is that there are many entry points to the system. Each 'node' of the system has to be kept physically and logically secured.
• Encryption Keys
• Encryption works by parts of the system exchanging secret keys that are then used to encrypt network traffic. So the more these keys are spread around, the more likely they are to be compromised.
• Corrupted Node
• If one node is compromised by a virus or hacking attempt then the rest of the system is vulnerable as well.

Pros and Cons of a distributed database

• Advantages:
• Resilient: A problem is one part of the organisation will not stop other branches from working
• Security: Staff access can be limited to only their portion of the database
• Network traffic is reduced so reducing bandwidth costs
• Local database still works even if the company network is temporarily broken
• Scaling: If a new store or branch is opened then it is straight forward to add an extra node to the distributed database
• High Performance: Queries and updates are largely local so there is no network bottleneck
• Easier to keep errors local rather than the entire organisation being affected
• Disadvantages:
• Complexity: A distributed database is more complicated to set up and maintain compared to a central database
• Security: There are many remote entry points to the system compared to a central database
• Data Integrity: More complex to make sure data and indexes are not corrupted
• Data needs to be carefully partitioned to make the system as efficient as possible
• Not so efficient if there is heavy interaction between branches, in which case a central database is the better option.