With the Issues Packet Loss Can Create on the WAN, Mitigation is a Priority

October 30, 2019

Network packets, the protocol data units (PDUs) of the network layer, are often taken for granted by network pros. We all get the concept: to transmit data over a TCP/IP network like the Internet requires the data be broken down into small packets (usually less than 1500 bytes) containing the relevant application data (“payload”) and headers. Routers forward these packets from source to destination and data encapsulation enables the data to traverse the TCP/IP stack.
The problem arises when this process fails, and packet loss occurs. Packet loss is, intuitively, when some packets fail to reach their destination.

Left unchecked, packets not reaching their destination can quickly become a major problem in an enterprise. When apps demand real-time data streams, even a relatively small amount of loss can create major problems. For example, Skype for Business connections MUST keep packet loss under 10% for any 200-millisecond interval and under 1% for any 15-second interval. That’s not much room for error, and similar requirements exist for other mission-critical VoIP (Voice over Internet Protocol) and telepresence app, making packet loss mitigation an enterprise priority.

Let’s explore packet loss in more depth and explain how Cato can reduce it on the enterprise WAN.

How Much is Too Much?

When discussing WAN optimization, the question of “what is an acceptable level of packet loss?” comes up quite a bit. I’m not a big fan of labeling any level of packet loss as “acceptable”, although a dropped packet here or there isn’t a major concern. As a rule of thumb, random packet loss exceeding about 1% can noticeably degrade the quality of VoIP or video calls. As packet loss increases, calls get choppy and robotic, video cuts in and out, and eventually connections are lost.

The surge in UCaaS (Unified Communications as a Service) popularity adds another wrinkle to the problem of packet loss. With voice and video services residing in the cloud, enterprises need a predictable low-latency connection to UCaaS providers like RingCentral, 8×8, and Telstra. In many cases, the public Internet is too unreliable for the job and MPLS (Multiprotocol Label Switching) is too inflexible and expensive. In addition to packet loss – latency, jitter, and security also become a concern with UCaaS. We deep dive on this topic in 4 Ways Cato is Perfect for UCaaS.

Detecting Packet Loss

Packet loss is calculated by measuring the ratio of lost packets to total packets sent. For example, in the ping output below, we see 1/5 of our packets did not make it to catonetworks.com, for a total of 20% packet loss.
ping catonetworks.com -t

Pinging catonetworks.com [203.0.113.2] with 32 bytes of data:
Reply from 203.0.113.2: bytes=32 time=105ms TTL=56
Reply from 203.0.113.2: bytes=32 time=136ms TTL=56
Reply from 203.0.113.2: bytes=32 time=789ms TTL=56
Reply from 203.0.113.2: bytes=32 time=410ms TTL=56
Request timed out.

Ping statistics for 203.0.113.2:
Packets: Sent = 5, Received = 4, Lost = 1 (20% loss),
Approximate round trip times in milli-seconds:
Minimum = 105ms, Maximum = 789ms, Average = 360ms

Tools commonly used to detect packet loss include:

ping. This is the simplest tool to detect packet loss and can be effective for ad-hoc troubleshooting. However, since many firewalls block ICMP (Internet Control Message Protocol) and it has a low priority, ping isn’t always enough.

tracert/traceroute. tracert (Windows) and traceroute (*nix) help identify the specific hop where packet loss begins.

Network monitoring software. Software applications like SolarWinds Network Performance Monitor, PRTG, Nagios, and Zabbix can all help monitor for packet loss at scale. For enterprise WAN, Cato Cloud’s Intelligent Last-Mile Management (ILMM) continuously measures packet loss in the last-mile.

Causes of Packet Loss

Detecting packet loss is one thing, but knowing how to identify the root cause is another. Common causes of packet loss include:

  • Routers with heavy CPU load. Routers have a finite amount of compute capacity, if the CPU load gets too heavy, packets can be dropped.
  • Security breaches. Malware or Denial of Service (DoS) attacks can consume a significant amount of bandwidth and resources, leading to packet loss.
  • Misconfigurations. Oftentimes, the cause of network outages is human error. The same holds true for packet loss. Misconfigured switches, routers, servers, or firewalls can lead to dropped packets. A textbook example is using half-duplex where full-duplex is needed or vice-versa.
  • Network congestion. The more traffic there is on a network, the more likely packets are to be dropped before reaching their destination.
  • Faulty hardware. Bad cables, routers, servers, and switches can all lead to packet loss and intermittent connectivity.
  • Software bugs. Packet loss can be related to a bug in a given software or firmware and updating may fix the problem.

How Cato Cloud Mitigates Packet Loss for The Enterprise WAN (with proof!)

With all the potential causes of packet loss and the Quality of Experience (QOE) issues it can create on the WAN, mitigating it is a priority. Cato Cloud has a number of built-in features that makes the WAN resilient against packet loss, such as:

  • Forward Error Correction (FEC). Enables the correction of packet loss predictively without the need for retransmission, reducing network congestion.
  • Identity-aware Quality of Service (QoS). Identity-aware routing and business process QoS take standard QoS to the next level by allowing critical data (e.g. an executive call) to be prioritized over standard traffic.
  • Dynamic Path Selection and Policy-based Routing (PbR). By proactively working around brownouts and blackouts, the Cato network automatically ensures packets are routed over an optimal path every time.
  • Active-active link usage. Ensures performance degradation in a single last-mile link can be overcome.
  • Packet duplication and Fast Packet Recovery. Help ensure rapid and reliable delivery of packets to reduce last-mile packet loss.

Just how effective is Cato at mitigating the effects of packet loss? RingCentral conducted testing that demonstrated Cato delivers high-quality voice connectivity across connections with packet loss up to 15%. If you find it hard to believe, check out this webinar and hear it for yourself.

Many Cato users have already experienced these benefits first hand. For example, according to Alewijnse ICT Manager Willem-Jan Herckenrath, when comparing Cato to MPLS, “Latency and packet loss are low. Even the users outside of Europe have the same or better user experience with our HD video conferencing and our CAD system (which runs over Citrix)”.

If you’re interested in learning more about how Cato can reduce packet loss on the enterprise WAN, contact us today.

Dave Greenfield

Dave Greenfield

Dave Greenfield is a veteran of IT industry. He’s spent more than 20 years as an award-winning journalist and independent technology consultant. Today, he serves as a secure networking evangelist for Cato Networks.