Management BYOD Infrastructure IoT Storage Security Privacy

Current Filter: Network>>>>>Feature>

PREVIOUS

Filtered Articles:3 of 237   Current Article ID:6378

NEXT



Network diversity

Editorial Type: Feature     Date: 01-2016    Views: 1308   







Dual infrastructure and diverse fibre links help to increase redundancy. However thorough evaluation and careful management are essential. John Volanthen, Technical Director at Hub Network Services explains

Locating servers in a modern tier 3/4 data centre can be the first step to securing data assets and increasing their availability. However, with the cost of data centre space and communications rapidly falling, organisations are recognising that operating dual infrastructure, potentially located in different cities or continents, is increasingly an affordable best practice for improving redundancy and security.

For this to succeed, developers and network engineers must understand each other's needs and limitations. An application that works perfectly on a system housed within a single location may fail spectacularly under geographical diversity because a few milliseconds latency has been introduced. Much greater emphasis is required on testing, especially under full load.

Understanding the true cost of downtime or data loss is the first stage of any geo-diversity project. This should inform the budget and project scope. Factors to consider are the overall objectives, the true cost of downtime and data loss, the acceptable timescale to restore and recover data, and the human capital involved.

A key design decision, potentially with wide implications, will be the configuration of the sites: especially choices around active/active (both operating simultaneously) versus active/passive (single operating site with the other mirroring).

The advantage of active/active is that both sites are known to work, whereas within an active/passive environment it is all too common for backup solutions (passive site) to remain untested. Best practice demands that any backup solution should be regularly and rigorously tested, for instance under expected load. A backup solution which an IT department is too frightened to test and use is not worth having.

Keeping both sites synchronised is critical to maintain data integrity. Some methods for handling this follow.

Synchronous replication: Distance is critical due to latency requirements as both primary and secondary sites must confirm that data has been written before an application can proceed. Transmission distances greater than a few kilometres can cause considerable performance problems.

Asynchronous replication: This option works over any distance, but increases the risk of data loss. This can be minimised using a journaling system, similar to that which is used on the Linux file system, ext3. This ensures that data writes can be rolled back to a known point.

Snap shots: Here data is copied in a state at a specific time. This is particularly applicable for large virtual machine images. Tools such as VMware or Veeam are ideal.

Another significant challenge for delivering geo-diverse Internet facing applications involves presenting IP addresses at multiple, geographically separate locations. Should the two sites both become active simultaneously then significant data corruption can result. Care also needs to be taken to ensure that traffic takes an optimal path. The increase in the volume of traffic caused by sub-optimal routing significantly increases the latency, the size of WAN needed and thus the expense of the system. An example would be an http: request traversing an inter site link many times.

Though all of these considerations are essential, they are secondary to the importance of selecting stable and robust fibre links between locations that are known to be truly diverse. To mitigate failure outside of the data centre, be sure to thoroughly evaluate which fibre routes and providers' best serve your needs. These are complex to unravel and external guidance may help.

Furthermore, it is imperative that each site is incorporated into a functional ring system and hardware clusters have sufficient distance separating them to negate local complications. As highlighted by the recent flooding in the North of England, for those companies affected, a secondary cluster in the less sodden South of the country would have greatly reduced their risk of downtime. NC

Like this article? Click here to get the Newsletter and Magazine Free!

Email The Editor!         OR         Forward ArticleGo Top


PREVIOUS

                    


NEXT