Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

WAN optimization: Network vs. application layer optimization

Larry Cormier, SVP of marketing for Silver Peak Systems | Dec. 6, 2011
The application layer of the Open System Interconnection (OSI) stack has long been regarded as a focal point for wide area network (WAN) optimization. Efforts to optimize application delivery were centered here because the pain was rooted in optimizing only a handful of specific applications at branch or remote offices.

This leads us to the network-centric WAN optimizer, which does not care if applications change or get updated. If we look back at our OSI stack, network-centric WAN optimization is handled much lower, at the network layer or Layer 3. This means no plug-ins are required for optimization, and the assurance that any new applications added to the environment are optimized using the existing investment in WAN optimization technology.

Consider for example an average enterprise that can have as many as 80 applications -- everything from ERP, to unified communications, to backup and data replication. Some of these applications might be purchased from third-party software vendors, some might be hosted in the cloud and others may be developed internally. If the enterprise is to depend on the application-centric WAN optimization approach, the time, cost and effort to ensure optimization is working properly for each application could be astronomical, and quite possibly unmanageable.

Quality control matters with MPLS and Shared WANs

It is not just applications to consider when evaluating WAN optimizers, but also the network itself. MPLS and shared networks are growing in popularity because of the low cost and flexibility. However, these popular WANs are plagued with network quality issues such as packet loss and out-of-order packets that adversely affect applications.

This is another area where network-centric WAN optimizers have an advantage, primarily because they are just that, "network-centric." Advanced technology such as Forward Error Correction (FEC) can only operate effectively at the network layer to improve the performance of applications traversing the WAN. And this is another area where the application-centric WAN optimizers differ.

Network WAN optimization can reconstitute lost packets in real-time to avoid delays that would otherwise come with multiple round-trip retransmissions. This proactively fixes the underlying network issues before the application is impacted. Conversely, application-based WAN optimizers typically must retransmit lost or out-of-order packets because it operates much higher in the OSI stack, Layer 7. This ultimately creates additional unnecessary traffic on an already constrained WAN.

Scalability is also a factor to consider, and an area where WAN optimizers differ. For example, virtual desktop infrastructures (VDI) generate many flows, and for hosted VDI deployments, WAN optimization flow count support is critical. More VMs equals more application usage, which translates to more required flow, protocol, and transaction support. Network-centric WAN optimization is built to support large flow counts, resulting in increased reliability and performance for emerging applications like VDI.

Of all the reasons for performing WAN optimization at the network layer, application independence is probably the most obvious. A network-centric WAN optimization approach enables organizations to future-proof their networks and allows for customers to get a better handle on trends such as cloud computing, virtualization and data center consolidation.


Previous Page  1  2  3  Next Page 

Sign up for Computerworld eNewsletters.