Symmetric Optimization in the Cloud

There are some issues, however, that will persist when making a cloud deployment. One of them is an increased usage of your Wide Area Network (WAN) connection. Whether you are going to a private cloud in a remote data center or a public cloud in someone else’s data center, all communications with applications will have to flow through your WAN.

The WAN connection is slower and more prone to packet loss and latency, and data transferred over it needs to be protected from man-in-the-middle attacks. These are issues that all WAN communications have faced, almost since the inception of Wide Area Networking, and companies created solutions to minimize the risks. WAN optimization is the broad name attached to these solutions. By putting a specially designed system at each end of a connection, packets on the way out can undergo compression, TCP optimization, de-duplication, and encryption,; and at the destination, all of these things can be unwound, resulting in the same exact thing coming out that went in. The improvements for WAN connections can be truly astounding, depending upon the quality of the connection and the traffic being sent through it.

In the case of WAN optimizations, de-duplication on-the-wire is used for larger chunks of repeating data, while compression is used for what is left. TCP optimizations attempt to mitigate latency and packet loss by imposing rules about things like retries that are part of the TCP standard. Taken together they are a powerful set of optimization tools that can speed communications over the WAN many times over.

Not in the Cloud You Don’t!

But cloud environments do not generally offer WAN optimization, and even if you find one that does, WAN optimization techniques and procedures are unique to vendors – meaning the cloud provider would have to use the same WAN optimization gear as your organization uses. That’s unlikely, and again, it is also highly unlikely you would find a cloud provider that even offered access to WAN optimization hardware unless you negotiated for it at contracting time.

And this situation has created an environment where stand-alone applications that do not need access to the data center in any way are the only ones considered suitable for cloud deployment – particularly in public cloud scenarios.

But there are many applications that could make use of the cloud for reasons of traffic patterns or volume requirements, but need access to internal data center resources. In fact, it could be argued that the truly independent application is a rare beast in the modern data center, and even if one is deployed, it will eventually need integration into the core data center for reporting or security or accounting purposes. This is one of the things that keeps the number of applications actually deployed to the cloud for production use low, though the potential is very high.

Virtualize It.

That is where virtual WAN optimization enters the picture. Virtual WAN optimization, either standalone or as part of a virtual application delivery controller, can be deployed anywhere that you can deploy a virtualized machine, which is most cloud providers. That offers an endpoint “in the cloud” that can handle the cloud side of data reduction and encryption operations.
With a virtual WAN optimization controller in place on the cloud side, the standard WAN optimization techniques that have been used between data centers and across extranets can be applied to cloud implementations. While high-bandwidth operations to the cloud like VM migration or replication definitely benefit more from this scenario than lower-usage applications, all applications communicating between the cloud and the data center will benefit from the optimizations.

There’s Money in It for You Too.

The other benefit of virtual WAN optimization controllers is a cost savings. Most cloud vendors charge you for the volume of data you transfer in and out of their data center. While these fees are generally small, because they are charged every month they will accumulate. A WAN optimization controller running inside the cloud receives compressed and de-duplicated data that is many times smaller than the source data (or the size it will be after it is reconstituted at the virtual WOC). Let us assume that your implementation reduces the size of data on-the-wire to 20% of actual unreduced data size. That is 1/5th the transfer amount you would have to pay for each month; even if the dollar amounts are small, over the life of your relationship with the cloud vendor they will add up to respectable sums of money.

As of the writing of this article, Amazon EC2 charges twelve cents per gigabyte after the first gigabyte each month – assuming less than 10 Terabytes of transfer in the month. For a popular application or data replication that communicates back and forth with the datacenter at the rate of 1 terabyte per month, the cost without a virtual WOC would be 1023 * $0.12 = $122.76 per month. With a virtual WOC environment that actually reduces throughput to 20%, that would bill out as (1024 Gigabytes * 20% – the first gigabyte * $0.12) less than $25 per month.

That is not an insignificant savings, even over a single year’s contract.

Conclusion

WAN optimization has been around and improving for many years. With the introduction of virtual WAN optimization controllers and virtual application delivery controllers with WAN optimization functionality built-in, functionality that improves communication between two physical locales can now benefit communications between a physical locale and the cloud also. Given the added benefit of data reduction going out of the data center through the WAN connection and less data going into the cloud connection, there’s compelling reason to make use of a virtual WAN optimization controller in the cloud.