What is re­dund­ancy in computer science?

When operating relevant IT systems and handling critical data sets, in­ten­tion­al re­dund­ancy should be an integral part of danger pre­ven­tion and data pro­tec­tion. Systems are redundant when the same technical com­pon­ents and data sets are available several times in parallel, pro­tect­ing against loss and failure. Re­dund­ancy does not only have ad­vant­ages in terms of storage capacity and IT equipment, however.

The noun ‘re­dund­ancy’ or the adjective ‘redundant’ and its meaning go back to the Latin ‘redundare’. It is composed of ‘re’ for ‘back’ and ‘unda’ for ‘wave’, and it denotes something that exists in abundance, i.e., also something that exists multiple times and in excess. While ‘redundant’ is barely found in everyday language, so-called re­dund­ancy plays a far greater role in computer science.

When we talk about re­dund­ancy in IT, we are talking about data centre security and system avail­ab­il­ity. It refers to data and system com­pon­ents that are numerous, parallel, du­plic­ated or mirrored and are thus available in abundance. Depending on the context, this can have both positive and negative con­nota­tions in IT. Re­dund­ancy in the positive sense stands for system-critical data sets that exist multiple times or are dis­trib­uted across multiple servers. Re­dund­ancy, when neg­at­ively connoted, is found in the form of un­in­ten­tion­al data du­plic­ates that occupy storage space.

First, then, a dis­tinc­tion must be made between un­in­ten­ded and intended re­dund­ancy:

Ac­ci­dent­al or un­in­ten­tion­al re­dund­ancy in an or­gan­isa­tion’s systems or data centres is found in the form of multiple data sets stored in one site or dis­trib­uted across multiple sites. Data du­plic­a­tion makes main­tain­ing records more com­plic­ated and leads to data anomalies. In­con­sist­en­cies are the result as it is not clear which data should be accessed or which records are current. In addition, un­in­ten­tion­al re­dund­an­cies occupy important storage space and consume energy un­ne­ces­sar­ily. This is prevented by nor­m­al­isa­tion of databases.

Intended re­dund­ancy means the planned, repeated design of technical com­pon­ents to secure a server, strengthen supply paths and protect system/company-critical data. A dis­tinc­tion can be made between the following concepts in re­dund­ancy:

• Func­tion­al re­dund­ancy: Multiple and/or parallel technical system com­pon­ents mostly within one plant.
• Geore­dund­ancy: Locally separated, multiple data centres or data sets.
• Data re­dund­ancy: Multiple backed up, mirrored or parallel data sets

If hardware damage, system failures or cy­ber­at­tacks occur, companies can protect them­selves against data loss and the failure of business-critical processes through re­dund­ancy. Data is stored multiple times, and con­sist­ently in different locations, while important com­pon­ents such as supply routes for energy and air con­di­tion­ing are at least du­plic­ated.

Depending on the com­pon­ents installed, a dis­tinc­tion is made between:

• Ho­mo­gen­eous re­dund­ancy: Repeated design of com­pon­ents that are identical in terms of man­u­fac­turer tech­no­logy. The dis­ad­vant­age of this is that the sim­il­ar­ity of the com­pon­ents means that the risk of an overall failure due to man­u­fac­turer errors or attacks directed at the man­u­fac­turer remains high.
• Diverse re­dund­ancy: Repeated design of com­pon­ents that differ by man­u­fac­turer, function, and type, making general system failures, uniform wear, and man­u­fac­turer-related failures less likely.

Re­dund­ancy as a criterion for system prop­er­ties and system security exists in the following forms:

• Redundant technical com­pon­ents: System com­pon­ents of computers and networks, such as air con­di­tion­ing, supply lines and servers, which take over the tasks of failed systems and com­pon­ents through multiple design or serve as a backup; this applies both to technical com­pon­ents within a system and to entire data centres that are available multiple times at different locations through geo-re­dund­ancy.
• Redundant in­form­a­tion: Su­per­flu­ous, un­ne­ces­sary, obsolete, or duplicate records that are not relevant to the system and usually occupy storage space un­in­ten­tion­ally.
• Redundant data: Repeated, mirrored, or dis­trib­uted data sets across multiple sites and servers that prevent complete data loss in the event of failure or damage through RAID cap­ab­il­it­ies, backups, vir­tu­al­isa­tion, or mirroring; serve as backup or disaster recovery; or provide faster access at a distance.
• Redundant bits: Added as extra bits during data transfers to prevent data loss during the transfer.

If it is important for companies to be able to access their servers without in­ter­rup­tion, they will implement a computer network. This consists of redundant servers made of cluster systems with several nodes. In a network, each as­so­ci­ated computer has equal access to existing databases and can take over access functions to critical data and ap­plic­a­tions for failed servers in the event of an emergency. To this end, a fail-safe is almost entirely guar­an­teed. Op­er­a­tion­al upkeep, launching without physical hard disks using network storage, and the re­place­ment of faulty servers can all be achieved without in­ter­rupt­ing business processes thanks to dis­trib­uted computer ca­pa­cit­ies.

Depending on the concept, redundant servers can be divided into two modes:

• Active/active cluster (symmetric): In the context of active/active clusters, servers operate in live mode as cluster nodes in which multiple computers work in parallel with dis­trib­uted power or in­de­pend­ently of each other. In the event of a failure, the computing capacity is dis­trib­uted to other servers in the network.
• Active/passive cluster (asym­met­ric): Active/passive clusters are called failover clusters and stand for the presence of redundant servers or network services that are in standby mode as a backup system and, by way of a switch-over, assume the functions of the main system in the event of a one-sided failure. This is automated by cluster manager/load balancer software. This also enables op­er­a­tion­al main­ten­ance without any loss in per­form­ance.

To implement re­dund­ancy in network systems and computing fa­cil­it­ies, there are various forms and concepts:

RAID stands for ‘Redundant Array of In­de­pend­ent Disks’ and refers to various physical storage devices (RAID arrays) that are joined together to form a partition. In this way, the con­sist­ency and integrity of data sets is main­tained even where an error occurs, and com­pon­ents can be exchanged without loss. This is done, among other things, by mirroring hard disks or by using parities with dis­trib­uted data in the array. However, RAID systems should always be used in con­junc­tion with an in­de­pend­ent backup of all critical data.

As already shown with the example of active/active or active/passive clusters, a computer network as a high-avail­ab­il­ity or load-balancing cluster offers greater avail­ab­il­ity, load dis­tri­bu­tion and fail-safety through elim­in­ated single points of failure and con­tinu­ous processes.

Geore­dund­ancy is often found as a re­dund­ancy concept in the clus­ter­ing of computing systems. It is used when par­tic­u­larly critical systems need to be protected against failure. In this case, identical computing systems are con­struc­ted locally and sep­ar­ately from each other, and data, geo­graph­ic­ally, is stored in­de­pend­ently. If one data centre fails, the redundant data centre can assume each task or com­pletely recover the data sets. Optimal geo-re­dund­ancy exists when ad­di­tion­al data backups are available in further fa­cil­it­ies.

Snapshots are virtual images or snapshots of hard disks and enable data and system con­di­tions to be backed up re­dund­antly in other storage centres. In the event of data loss at one location, data recovery can be im­ple­men­ted, therefore. Storage re­quire­ments for snapshots are sig­ni­fic­antly lower than for data copies because they are reference markers for data storage locations and not actual copies.

A backup requires more storage volume compared to snapshots as the data is copied and stored as a backup copy and in redundant form. Thanks to this data re­dund­ancy, the complete data set can be restored. Even with a redundant computer network, an ad­di­tion­al backup is always re­com­men­ded.

With CDP, data is stored as part of a con­tinu­ous backup, which monitors changes and auto­mat­ic­ally updates them in the backup. It is therefore a data re­dund­ancy that backs up critical data in real time and protects against failures.

The ad­vant­ages of in­ten­tion­al re­dund­ancy are obvious: Systems and networks that have multiple technical com­pon­ents and storage devices offer greater re­si­li­ence, faster data access and more sus­tain­able op­er­a­tions. Data recovery and con­tinu­ity are ensured even in the event of severe failures. The dis­ad­vant­age of redundant systems is the re­l­at­ively high cost of several com­pon­ents, required storage space, and for the con­tinu­ous updating of data copies.

Nev­er­the­less, re­dund­ancy in the data centre is becoming in­creas­ingly important in view of new cyber threats, outdated system tech­no­logy and strict data pro­tec­tion re­quire­ments. Both end users and data centre operators should take care not only to integrate failover and system security ef­fi­ciently through redundant concepts, but also to mark them as a com­pet­it­ive advantage and USP through data centre tiers.