ARN

WAN optimization: Network vs. application layer optimization

This vendor-written tech primer has been edited by Network World to eliminate product promotion, but readers should note it will likely favor the submitter's approach.

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.

But the problem has become broader. There are too many applications, platforms and versions of applications that need to be accounted for, and the list continues to grow. It has become clear we require a different approach. IT organizations must not only deploy WAN optimization to recoup valuable bandwidth for specific applications, but must deploy a solution that cost-effectively optimizes the entire application environment now and in the future, and do it without increasing network infrastructure complexities.

TECH ARGUMENT: WAN optimization hardware vs. WAN optimization services

The network experts responsible for the fast, reliable and secure delivery of applications across the WAN understand the need for optimization tools that supports all traffic types, not just certain types. They want complete control and visibility into the business-critical applications as well as the not-so-critical apps. The point is that every application is, or should be, present in the enterprise for a reason, and it's important to realize that each has the potential of adversely affecting the network.

To understand this, one needs to look at underlying protocol support. While the lion's share of network "pain" has traditionally been with TCP applications, the number of business-critical applications that are non-TCP based is growing. Voice and video come to mind, as well as desktop virtualization (VDI). If a particular WAN optimizer doesn't support the non-TCP applications, it simply bypasses that traffic. This alone should set off red flags for any network professional.

Then there are other applications like disaster recovery, storage virtualization and data replication where breakthrough technologies like EMC VPLEX go beyond the TCP-based communication. There are also the growing number of proprietary protocols and custom applications many enterprises rely on to run their business, including customer enterprise resource planning (ERP) and communication applications.

For any of these situations, the network professional doesn't want to have to tell the WAN optimization vendor how the custom or proprietary apps work, they just want them to work.

For application-centric WAN optimizers to solve today's emerging challenges, the answer isn't as simple as it used to be. Today, these vendors need to apply "plug-ins" for every application the customer wants to optimize. This is not only complex given the ever-growing set of applications, but also presents challenges with regard to the dynamic nature of the applications. The most common example: If the software or application vendor makes a change to the app, the application-centric WAN optimization vendor most update the plug-in to support it.

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.

Network-based WAN optimizers provide the advantage of optimizing all current and future enterprise applications. Whether enterprises update to new versions of their applications or update existing solutions, by default, network WAN optimizers continue to operate in lock-step with the dynamic application environment.

While there are a number of key return-on-investment (ROI) drivers for deploying WAN optimization, including number of locations, total size of the WAN and the nature of the traffic, being able to holistically address all of these concerns at the network layer means they can be addressed by a single appliance.

Larry Cormier is SVP of marketing for Silver Peak Systems.

Silver Peak is a privately held company in Santa Clara, Calif., that improves backup, replication and recovery between data centers and facilitates branch office server and storage centralization by improving application performance across the Wide Area Network (WAN). Visit www.silver-peak.com for more information.

Read more about lan and wan in Network World's LAN & WAN section.