Reliability and Quality of Service

The OnSIP Hosted PBX service is delivered across the Internet and are, by definition, "best effort". While we have gone to great lengths to architect our system and interconnections such that the components we do control are completely redundant and perform optimally at all times, our service also depends on the performance of external components such as local power, broadband carriers, and Internet backbones.

Some broadband connections may have less than desirable quality. When IP packets are lost or greatly delayed at any point in the network between users, there will be a momentary drop-out of communication. This is more noticeable in highly congested networks. Typically, when we do see either packet loss or high latency, it is nearly always in the last mile. Typically this is easily uncovered and easily rectified by the broadband carrier.


End User Devices and User Agents

We cannot stress enough that in our experience the single greatest factor affecting the end user's qualitative experience is the device they are using. IP phone software and hardware varies greatly and we strongly recommend investing in quality devices.


Packet Loss, Jitter, and Latency

  • Packet Loss can be caused by too much traffic on the network resulting in packets being dropped, incorrectly configured routers, hardware failures, as well as a host of other issues. Whatever the cause, lost packets result in gaps and stutters in communication and can greatly affect the perceived quality of communication. When a packet is lost the associated communications data simply disappears.
  • Jitter is variation in packet latency. The effects of jitter can be mitigated by storing voice packets in a jitter buffer upon arrival and before producing audio, although this increases delay. Furthermore, we have found that the quality of jitter buffer implementations varies greatly across devices. Regardless, high amounts of jitter can result in communication disruption and call quality issues, but generally to a lesser extent than packet loss.
  • Latency is the time from the source sending a packet to the destination receiving it. As long as packet loss and jitter are avoided, latency tends not to be an issue unless it is extreme. For example, the 100ms (0.1 second) round trip latency on a call from California to New York is unlikely to be noticed by the callers, but dropping 1% of the packets on any call will likely be noticeable regardless of the latency. That is, the consistency of packet arrival tends to be much more important than the time it takes a packet to arrive.


While we have limited control over what happens to packets outside our network, OnSIP has engineered its system, network, and Internet connectivity to minimize packet loss, jitter, and latency. At all times, by design, we operate at 0% packet loss and less than 1ms of latency across our network and across all our connections to the Internet.

Secondly, we currently employ three redundant data centers. Two data centers are in New York City and the third is located in Los Angeles.


Prioritization, Type of Service, and Differentiated Services

Fixed delays may not be controllable, but some delays can be minimized by marking packets as being delay-sensitive. While we do currently set the Type of Service (ToS) on delay sensitive IP packets exiting our network, we have found that some Internet service providers will reset it upon ingress to their network so the setting is lost by the time the packets arrive at their destination. Furthermore, unless the end user's ISP is prioritizing packets based on this information, it is of little or no utility as far as QoS is concerned.

Nonetheless, we currently set the IP ToS byte defined by RFC 791 on all delay sensitive datagrams to either 0x18 (Low Delay and High Throughput) or 0x10 (Low Delay). Defined in RFC 2474 and RFC 2475, the Differentiated Services (DiffServ) standard supersedes the original specification for defining packet priority described in RFC 791. DiffServ increases the number of definable priority levels by reallocating bits of an IP packet for priority marking. We are planning on implementing DiffServ in the near future such that delay sensitive packets will be classified for Expedited Forwarding (EF). The EF DiffServ also contains the ToS Low Delay bit, which is treated by some routers without any configuration. This is sometimes called Priority First In First Out (PFIFO), or First In First Out (FIFO) with Priority.

While there may be little QoS benefit gained from our efforts to prioritize packets bound for the Internet, many of our users have improved the quality of their experience by implementing some packet prioritization locally. This can be particularly effective when the customer has an asymmetric Internet connection where the downstream bandwidth is significantly greater than the upstream bandwidth (typically the case for DSL and cable modem based Internet connections). In these cases prioritizing packets heading upstream to the Internet can provide tangible improvements in QoS since the low bandwidth upstream channel is likely to be the main (or only) point of congestion. Currently, many IP phones and broadband routers come default from the factory doing DiffServ based prioritization of upstream packets and there is little, if any, configuration that needs to be done by the end user to gain this QoS benefit.


Network Alerts

If there is ever an issue with our service or with one of our upstream carriers, we post details at You can view the history of these alerts here.

An alert is automatically posted to the administrative login page. Additionally we publish Twitter posts to keep our customers informed.

Was this article helpful?
0 out of 0 found this helpful