With data center IP traffic growing at a 25% annual rate to a projected 7.7 zettabytes per year by 2016, the obvious question is what’s driving all this traffic? Cloud-based file sharing? Hybrid cloud backup and archiving? Higher quality video conferencing? Viral videos of goats and puppies?
There’s no doubt that there’s broad user demand for data for any number of use cases; but that drives only 17% of the IP data center traffic. That 17% is the very visible tip of the iceberg – it’s very apparent, but the bulk of data center IP traffic is occurring “below the waterline.”
Inter-data center traffic is 7% of IP data center traffic. An example of inter-data center traffic is replication between AWS (Amazon Web Services) regions. This type of traffic is necessary to ensure high availability for a multi-data center cloud offering – the type that most major club providers (e.g., AWS, Google, Microsoft Azure, Rackspace, etc.) offer today. Software/system updates across data centers is another major cause of this type of traffic.
The vast majority of data center IP traffic occurs within the data center itself. What does this mean? This is the traffic that occurs due to storage, production, and development data within the data center itself. A typical example of this type of data traffic may be found in AWS among the “offline” (4 hour restore SLA) lowest-priced object-based Glacier storage, the online relatively higher latency S3 (Simple Storage Service), and the online and lower latency EBS (Elastic Block Store) – this type of traffic occurs when data is transferred among these different storage types.
Within these data centers, what about the impact of cloud workloads versus traditional data center workloads? This will be discussed in part 4 of this series.