bizibly

How Aryaka Optimizes WAN Traffic with Fewer Application Acceleration Proxies

In my last blog, I discussed why application proxies are becoming less frequently used in the WAN Optimization space today.

To recap, the NUMBER ONE reason that technology has largely moved away from the “Layer-7 Proxy” model is because of the high propensity to “break” many of the applications they purport to help.

The Aryaka Approach

App_Proxy

So how can Aryaka optimize traffic without dozens of Application specific Layer-7 proxies like “the other guys”?

Simple. By fixing the root cause of the application problem, rather than trying to mask the symptom.

For instance, look at data deduplication technology. While it is true that old “block based” versions of deduplication engines needed to know where the data portion of an IP packet began and needed to be able to match a block of contiguous data in order to get reasonable deduplication results, the newer “byte level” deduplication engines, such as the technology that Aryaka uses in our WAN Optimization software stack, are able to dedupe data anywhere in the IP packet, making the need to have specific proxies for applications like NFS and SRDF unnecessary.

Lightweight, transactional applications like MS-RDP or Citrix ICA do not benefit from deduplication anyway. In these cases, the amount of data in the IP packets is very small, and is already compressed. The performance problems experienced by the end-users of these applications are generally caused by network related issues like high or varying network latencies, or packet loss.

There are, actually, only three major technical solutions for improving end-user performance problems when using these types of protocols over the WAN:

  1. Stabilize end to end latency.
    Varying latencies and latency spikes can result across most WAN provider networks due to routing changes, interim congestion points, bandwidth oversubscription, faulty equipment, etc.

    Because Aryaka owns the network, we have built in capacity and redundancy to ensure that there is always a stable network core. With predefined traffic routing, and multiple redundant paths, Aryaka can provide an application with a stable, consistent, and predictable network condition which improves the efficiency of application send and receive buffers and results in a much smoother interactive end-user experience.

  2. Reduce packet loss.
    Combined with varying latency rates, packet loss is the biggest culprit in causing performance problems for users of transactional applications. Lost packets cause lags and delays in things like mouse movements or onscreen typing, as well as issues with voice quality or image quality with collaborative programs like MS-Lync. While there are a number of different technologies out there that try to mitigate packet loss problems on lossy links, none are as effective at improving application performance as having a clean, low loss connection.

    Owning the network core allows Aryaka to remove the chief causes of packet loss, which typically results from network congestion or faulty equipment. Aryaka never oversubscribes their bandwidth so there is never core congestion. In fact, Aryaka maintains a significantly higher amount of bandwidth than which is guaranteed to our customers. This allows for quick provisioning of additional bandwidth to customers when needed, and allows for customer “bursting” about their provisioned rates. If packet loss were to occur on a WAN segment due to faulty equipment, the traffic is immediately and seamlessly transferred to a redundant WAN link to prevent impact on customer traffic.

  3. Remove congestion control responsibility from the application end point, and put it on the network where it belongs.
    The underlying transport protocol for many applications is TCP, and TCP asks the end user to try and “figure out” how available the intermediary network is for sending data, and then adjusting its transmission rate to fit the network. While this approach worked well 15 or 20 years ago when wide area networks were typically very low bandwidth and often congested, today, this actually results in the application becoming the bottleneck.

    While most WAN Optimization devices provide some measure of TCP congestion control for their appliance based point to point solutions, Aryaka only has to provide this particular feature at the edge, from the customer location to the first Aryaka Point of Presence. Across geographic regions, Aryaka controls all of the data transit across our core network, so we have the ability to use newer, more efficient protocol algorithms designed to be more effective on the larger, longer WAN links in use today.

Traditional WAN Optimization Vendors Lack the Network Piece
WAN Optimization technologies have no control over the intermediary network conditions and are stuck trying to overcome or mask problems occurring over the WAN transit links. For example, they can do nothing for the packet loss or varying network latencies conditions which cause problems for today’s lighter weight WAN applications like Citrix, MS-RDP, VoIP, etc. This leaves traditional WAN Optimization technology stuck, trying to “mitigate” the intermediary loss and latency problems that occur across the WAN, rather than being able to actually resolve them.

Doing “More with Less” Application Proxies
There are still a few application layer protocols, like older versions of CIFS or SMB, which benefit from having a specific proxy for optimization over the WAN, and of course Aryaka supports these corner cases today. But that list is constantly growing shorter. The truth is that most protocols are behaving better on the WAN as they evolve.

By reducing the dependence on specific application layer proxies, Aryaka’s technology can provide better optimization while improving reliability and removing potential points of failure.

Our byte-level deduplication and compression technology supports all non-encrypted traffic, independent of application-specific data formatting.

And our multi-segment, optimized, private network architecture provides better performance to globally distributed enterprises – with 0% of the headache associated with frequent upgrades, patches, and fixes.

About the author

Robert Meyer

Robert Meyer is the Director of Systems Engineering at Aryaka.