Once upon a time, data center WAN traffic was mainly about backup; hitting backup windows was the primary driver for optimizing those links, and compression was the main WAN optimization technique used to handle it. Over the last several years, however, the traffic on the data-center-to-data-center WAN has changed enough to consider different approaches to data center WAN optimization now.
For example, as more than half of workloads in the data center have been virtualized, traffic between data centers is increasingly about moving operating-system (OS) images. Replicating primary storage has been a key disaster recovery (DR) discipline for a while, and has only become more important because the majority of OS images are now part of that data stream. Beyond that, though, with the construction of private clouds atop server virtualization, there is now even a need to support live migration of virtual machines (VMs) across MAN and WAN distances. Live migration brings with it needs for high volume, low latency data transfers that are infrequent, unpredictable and extremely important when they happen. So, data center WAN optimization needs to better deal with sudden spikes in demand on priority data flows.
More on data center WAN optimization
A WAN managers guide to data center interconnect
How to improve disaster recovery with WAN optimization
Comparing data center disaster recovery WAN optimization options
Data center consolidation demands WAN optimization, network upgrades
Now, too, application-to-application traffic between data centers is becoming increasingly important and challenging. As service-oriented architectures (SOAs) have continued to proliferate across companies and even more systems, application-to-application traffic is increasing in both volume and criticality. The volume is increasing in all senses: There are more messages; the maximum message size is increasing; and there are more applications (i.e., endpoints in the flows of traffic between data centers) generating streams of messages to each other. The traffic tends to be “lumpy,” though, with some application messages measuring in bytes but others in megabytes. Unfortunately, this inconsistency can make application performance variable as well. The end-users’ perception of performance will be controlled by how these application-to-application flows perform. Even though the users do not participate directly in any of these flows, they all happen at the back end, so guaranteeing solid performance is important for usability and user satisfaction.
Inter-data-center WAN optimization therefore needs to recognize applications and their associated traffic and, granularly, accelerate the parts that matter most to end-user performance (which will vary by application type and end user).
And, of course, changes in the enterprise data landscape have begun to radically change the nature of that core job of data replication. The biggest change is the emergence of “big data” in the data center. Big data doesn’t just refer to the steady growth in storage consumption that IT has been living with for decades. Big data means massive changes in data flows with several key characteristics:
- Demand for storage is coming from parts of the business that have not generated much data in the past.
- Storage-usage patterns are dramatically different.
- The meaning and the utility of data may not be immediately obvious.
- Data volume is increasing by an order of magnitude or more.
When folks, like facilities, become major new consumers of storage capacity, or when the bulk of storage data shifts from traditional databases or office files to video and audio, or when the annual 50% to 150% growth in data consumption jumps to 500% to 1500% growth, a WAN link between data centers engineered with old assumptions in mind can be brought to its knees. Data center WAN optimization links need to have big data in mind and work to accelerate it.
Big-data, private-cloud and virtual enterprises all mean that data center WAN optimization has to adapt to the changes. It will need to be more flexible, application-aware and granular, and to be all that while managing unprecedented numbers of data flows and volumes of data.
Read the other articles in this series:
- Data center WAN optimization must relieve new traffic pressures
- WAN optimization for the data center interconnect
- Inter-data center optimization: How to accelerate WANs without optimizers