SD-WAN VS MPLS -- Battle of the last mile
While telecom operators are increasing their infrastructure investment and Internet performance has improved in many regions, most companies are still reluctant to completely replace their MPLS infrastructure with a more cost-effective Internet IPVPN. For saving more budget for MPLS deployment for business-critical applications, Internet IPVPN is usually used to carry non-critical applications, to improve the cost efficiency of enterprise WAN. The reason behind is the IT managers have relied on MPLS for last 20 years and they are very concerned about the network stability and network performance which MPLS is good at.
As you may know, for stable network performance, some SD-WAN service providers can provide a private core network similar to MPLS, but for flexibility, SD-WAN chooses internet for last mile (from customer site to SD-WAN core) connectivity, this is different with MPLS which use private network for last mile connectivity. Therefore, the deployment method of SD-WAN has been criticized by many people, saying that it is “not as stable as MPLS” because SD-WAN use internet as the last mile connectivity. And "stability" is the most critical consideration for IT managers to make network solution selection decisions.
It can be seen that internet give SD-WAN the advantage (flexibility) and also the disadvantage (Instability). How the SD-WAN service provider solve disadvantage of last mile internet connectivity has become a key capability of a SD-WAN service provider. Regarding the solution, on the one hand, from resource perspective, it selects high-quality BGP Internet as the ISP of SD-WAN POP, and on the other hand, it is the network optimization technology which different with each other.
In this article, we will focus on the network optimization technology of the AXESDN, which is the secret that AXESDN can provide super better application performance than MPLS based on internet connectivity of the last mile. It is also one of the main reasons why customers choose to use AXESDN SD-WAN as the enterprise WAN connection solution.
Last mile connectivity solution of AXESDN
AXESDN leverages the following key technologies to reduce network packet loss caused by last-mile Internet access, and the impact of latency jitter on application performance:
Redundancy: AER redundancy and link redundancy
Link selection: The system actively monitors the packet loss rate and latency of each link and selects the link with the best performance for application traffic.
Link aggregation: Sends traffic simultaneously through the primary link and the secondary link. If one of the links has packet loss, another link carries the same packet to the peer. This solution can significantly improve application performance when link packet loss.
TCP optimization: Using technologies of “TCP retransmission optimization”, “TCP ACK mechanism optimization”, “TCP congestion control optimization”, and “Real transparent TCP proxy” to make sure the best TCP transmission performance for applications.
We did a test to demonstrate the effects of the above technology deployment.
Test environment topology (network latency is 180ms)
Test case 1: [Without optimization] Bandwidth cap is 10Mbps, network packet loss rate 6%, Windows to Ubuntu file transfer (303MB)
Test result: Although the bandwidth cap is 10 Mbps, the bandwidth utilization of the application is extremely low due to packet loss. The maximum bandwidth of the file transfer is 547 Kbps, and the average bandwidth usage is 136 kbps. (Note: Speed in below gif is MB/s, and the speed in screenshot is Mbps.)
Test case 2: [Without optimization] Bandwidth cap is 10Mbps, no packet loss on the network, Windows to Ubuntu file transfer (303MB)
Test result: Although the bandwidth cap is 10Mbps, the limitation of the TCP window by the old TCP protocol stack causes the bandwidth utilization of the application to be related to the network delay. The higher the network latency, the lower the transmission speed, so in this environment, the maximum bandwidth occupied by the file transmission is 3Mbbps (less than 1/3 of the upper bandwidth limit), and the average bandwidth usage is 1.3Mbps. (Note: Speed in below gif is MB/s, and the speed in screenshot is Mbps.)
Test case3: [with optimization] bandwidth cap is 10Mbps, network packet loss rate of 6%, Windows to Ubuntu file transfer (303MB)
Test result: The maximum bandwidth for file transfer is 10Mbps, and the average bandwidth usage is 2.3Mbps. (Note: Speed in below gif is MB/s, and the speed in screenshot is Mbps.)
Summary of test results:
According to the above test results, it can be seen that the bandwidth utilization of case 3 (with optimization & link packet loss rate 6%) is higher than that of case 2 (without optimized & 0 package loss). This is the proof that AXESDN optimization technology can provide better application performance than MPLS.
For free PoC test, please send email to sales@axesdn.com