Multi-protocol label switching (MPLS) is suddenly a hot topic as companies look to improve network performance without investing too much money in new equipment or services.
Basically, the technology is designed to route network traffic around congestion and failures and therefore speed things along without any involvement from the user. Wireless service providers employ the technology because it improves quality of service (QoS), reduces delays and helps keep operations on track in terms of service level agreements (SLAs) with customers. It is for this reason that up to 90% of the companies surveyed earlier this year by Nemertes Research admitted using or evaluating the technology.
Requires Free Membership to View
SearchEnterpriseWAN.com members gain immediate and unlimited access to breaking industry news, best practices for designing and managing Wide Area Networks, WAN Security, and more -- all at no cost. Join me on SearchEnterpriseWAN.com today!
Kate Gerwig, Editorial Director
|  | ||||||||||||||||
| |||||||||||||||||
The 'dirty little secret' surrounding MPLS, however, is that the technology is not perfect and packet losses are very common during information transfers. These losses can range from as little as .1% to as much as one to two percent – not critical if you are dealing with generic email and low-level data, but it could be a problem in mission critical applications like healthcare.
Network switches and routers don't usually have the capability to detect dropped packets, let alone correct them, say the experts. In most cases, these dropped packets may show themselves as a slowdown in network performance, which can result from a number of reasons. So, users tend to accept and overlook the problem as a way of network life.
"Generally speaking, most people do not have the tools to go and troubleshoot and track down the cause of periodic applications performance problems," said Rick Tinsley, president & CEO of Silver Peak Systems, Inc., a maker of WAN optimization solutions.
Managing MPLS packet loss
The solution, in most cases, is to detect and correct minor MPLS packet losses in real-time, using a technique borrowed from the disk storage world called forward error correction. What this tactic does is look at every single byte and packet that is sent over the network and track the order by assigning sequence numbers. It also assigns an extra parity packet that can be used to reconstruct failed packets on the fly, explained Tinsley.
When a group of packets is transmitted and arrive at their destination, these error correction tools can determine which packet was dropped and then mathematically rebuild that packet – most often with 100% accuracy, he noted.
Of course, if multiple packets are dropped or corrupted, there may be no way to reconstruct the lost bytes, so the only alternative would be to retransmit the data and hope that MPLS performs a little better the second time around.
What can network administrators who have gone the MPLS route do to eliminate the lost packet problem? Nothing really, since a little nip in the flow here and there is, for the moment, inherent to MPLS. Companies like Silver Peak do provide forward error correction tools in their WAN acceleration solutions, and network service providers try to minimize the problem and guarantee not more than .1% packet loss per month.
This may be a tough guarantee to honor, though, if customers are not actively tracking packet losses on a minute-by-minute basis, said Tinsley. Once they do document losses, however, companies may be able to renegotiate their SLAs to account for minor blips in service.