Managed cloud service offers alternative to DIY or WAN optimization appliances.
If you are trying to improve global access to your applications, you have probably considered one of several solutions: stringing together your own private network, purchasing WAN optimization appliances, or using a managed cloud-based service provider. Figuring out the benefits of each solution isn’t easy and it is hard to test for variations in Internet connectivity, specific applications and other conditions.
But what if a vendor could show you exactly the benefit in a particular use case, so you could understand what they are delivering? That was the idea behind a request that we made of two vendors in the managed cloud space, Aryaka and Virtela. The former responded to our inquiry with a series of tests; Virtela declined to participate.
Aryaka has its own global private network where it tries to make transit and latencies more predictable than the general Internet. They have points of presence in dozens of cities around the world and connect them with multiple redundant links.
They use application-level proxy servers, data de-duplication and compression to accelerate access to various SaaS and other cloud-based services, along with better site-to-site connections. By having their own network, in theory they can avoid the congested public Internet peering points and deliver better applications performance.
So we asked them to set up these two testing scenarios and measure the resulting latencies:
- A user in Bangalore is accessing shared files in a data center in Milpitas, Calif., and a Web server running in the Amazon Web Services facility in Ashburn, Va., through the general Internet using standard IPSec VPN tunnels.
- The same Bangalore user will access the same hosts using the private Aryaka network.
For both sets of tests, they transferred two different 60MB files: one that was easily compressible and one that wasn’t. They also added some local network congestion with a packet loss simulator to see the effect of last-mile conditions on overall performance.
The non-compressible files accessed over the private Aryaka network were transferred from five to 20 times faster than those over the general Internet. As the local network became more congested, the disparity in transfer time increased, favoring the Aryaka connection (see the graphic below). And as they used compressible files, the benefits also increased, favoring the Aryaka connection even more so.
What are some key takeaways from these tests? First, as I mentioned, we didn’t supervise or conduct the tests, but after reviewing the test plan and results, we think they offer some interesting results.
We didn’t ask them to test their own network compared to a traditional pair of WAN optimization appliances: that is obviously the next step for anyone interested in this topic. The WAN optimization method could be better for point-to-point paths, which is what they are designed for. Where Aryaka and other managed services can play is in the mixture of cloud and LAN connections that are being accessed across the globe.
And you should certainly test across your own network conditions and particular applications, too. Obviously, the two apps we tried (a standard network file CIFS share and an HTTPS file transfer) are pretty plain vanilla: more complex applications such as CRM, databases, and Web applications can have different latencies and round trip packet flows that could change the results we show here.
Second, Aryaka didn’t test for congestion in the “middle mile” or that portion of the Internet in between the origin and destinations mentioned above. That congestion could clearly add more of a benefit to their private network path, which doesn’t have that congestion.
Third, your own situation could vary depending on the distance and connectivity between your network assets and the Aryaka points of presence. The further or more miserable your connection to their private network, the lower the benefits.
Strom is the founding editor-in-chief of Network Computing magazine and has written thousands of magazine articles and two books on various IT and networking topics. His blog can be found at strominator.com and you can follow him on Twitter @dstrom. He lives in St. Louis.
