Shall I place the replicas on different Data centers based distance (Data centers distance)? Is it possible? I want reduce replication with out affecting data file loss due to disaster. So I think to consider the distance as deciding factor. But there is no base paper for it. Shall I consider it or not? Please help me.
Yes you can. I hope you can find papers for such scenarios. Try searching in IEEE explorer.
In the Hispanic Caribbean rules of international banking indicates that the primary and alternate data centers must have a minimum separation distance of 40 km. This may vary by country, seek advise regulations and laws to get an idea of the legal requirements for an alternate site services. In addition, there are a set of best practices and recommendations, such as that brought the Federal Reserve after the 9/11 attacks but are aimed at the financial sector. The distance between the primary site and the alternate should be several kilometers or thousands of kilometers, depending on many factors: types of hazards (earthquakes, floods, political risks, wars, etc.), type of activity (financial, telecom, etc ) and critical level (how much money is lost for downtime).
Yes. You can use the distance as a constraint in a federated datacenter depending on the type of service you are dealing with. Geolocation technologies are already available to decide the nearest datacenter based on distance. You will also have to consider the network latency between the federated datacenters. Also consider that moving replicas around also affect the load/consolidation ratio of destination datacenters too.
I have included some links for more reading.
http://www.cloudbook.net/resources/stories/cloud-service-placement
http://www.cisco.com/c/en/us/products/collateral/ios-nx-os-software/ios-xr-software/white_paper_c11-694882.html
http://www.opendatacenteralliance.org/docs/Long_Distance_Workload_Migration_Rev1.0_b.pdf
Conference Paper Intelligent Placement of Datacenters for Internet Services
Article Dynamic placement of autonomic internet services
Article Service Placement in Network-aware Cloud Infrastructures
It depends on complex your scenario is. You can use the distance as a constraint in a federated datacenter depending on the type of service you are dealing with. Geolocation technologies are already available to decide the nearest datacenter based on distance. You will also have to consider the network latency between the federated datacenters. Also consider that moving replicas around also affect the load/consolidation ratio of destination datacenters too.
I have included some links for more reading.
http://www.cloudbook.net/resources/stories/cloud-service-placement
http://www.cisco.com/c/en/us/products/collateral/ios-nx-os-software/ios-xr-software/white_paper_c11-694882.html
http://www.opendatacenteralliance.org/docs/Long_Distance_Workload_Migration_Rev1.0_b.pdf
Conference Paper Intelligent Placement of Datacenters for Internet Services
Article Dynamic placement of autonomic internet services
Article Service Placement in Network-aware Cloud Infrastructures
look at the QoS from the SLA with provider to proper estimate netw latency
You need to take into consideration that the traffic QoS between two points could vary time to time. Therefore, you might need to select replica location dynamically while retrieving any content.
Yes, it is one of the best practice to replicate your data in today's world to avoid from data loss due to any disaster. Try distributing your files using HDFS, by default it replicates your data at three centers. And you've to decide the distance of your data centers.
- Badges
- Science method