Latency and Jitter

Latency describes the travel time of data packets in the network, jitter their variation. This article explains how both quantities arise, why VoIP and video are especially sensitive, and which measures let companies noticeably improve response times.

The video conference stalls, the phone call sounds choppy: too little bandwidth is rarely behind such disruptions. Usually latency and jitter are responsible, that is the travel time of the data packets and its variation. Both quantities determine whether applications feel fluid and whether real-time communication works reliably. Anyone connecting sites and cloud platforms over long distances should therefore take them just as seriously as the throughput of a line.

What are latency and jitter?

Latency is the time a data packet needs to travel from sender to receiver. It is measured in milliseconds (ms), in practice mostly as round trip time (RTT): the duration for the outbound and return path together, as a simple ping shows it. Interactive applications perform many such round trips for every action, for example to establish connections and run database queries. Small differences in the RTT therefore add up to clearly noticeable waiting times.

Jitter refers to the variation in latency. If packets arrive sometimes after 20 ms and sometimes after 80 ms, jitter is up to 60 ms. Applications can compensate for a constant delay with buffers, but hardly for a strongly varying one. Closely related is packet loss: if packets are lost along the way, they have to be retransmitted or are missing from the media stream. Both further aggravate the perceived delay.

How it happens

Several factors determine how long packets are in transit and how strongly their travel time varies:

Physical distance: In fiber optics, light travels around 200 kilometers per millisecond. Between Europe and East Asia, distance alone produces an RTT of well over 100 ms that no technology can optimize away.
Number of hops: Every router on the path accepts packets, processes them, and forwards them. The more intermediate stations, the higher the total delay and the greater the probability of variation.
Queuing and overload: Under high utilization, queues form in routers and switches. Packets wait for different lengths of time to be forwarded, and this is precisely where jitter arises, plus packet loss during load peaks.
Detours in routing: The public internet chooses paths according to the commercial agreements of the network operators involved. The shortest route is rarely guaranteed, and traffic to Asia sometimes runs across several continents.
Last mile and access technology: Wi-Fi, mobile networks, and congested home-office connections cause variable delays that appear in no backbone statistic.
Processing along the way: Firewalls and VPN gateways add compute time for encryption and inspection. If they are under-dimensioned, this becomes a bottleneck of its own.

Why it matters

Voice quality suffers noticeably when the delay per direction rises above about 150 ms. Conversation partners then unintentionally talk over each other.
Video conferences react sensitively to jitter, because image and sound fall out of sync and larger buffers increase the delay further.
Cloud applications with many small requests, for example virtual desktops or database access, scale their perceived speed directly with the RTT.
The throughput of TCP connections drops under high latency and packet loss even when plenty of bandwidth has been booked.
Real-time data from production and logistics lose value when they arrive late or in the wrong order.
Reliable commitments for voice and video quality require stable network values that can hardly be guaranteed on the public internet.

Typical scenarios

A branch office in Shanghai accesses central systems in Frankfurt. The connection runs over changing internet paths, video conferences break up regularly, and file access takes noticeably long.
A company moves its ERP to the cloud. The line is barely utilized, yet screens and reports feel sluggish, because every action requires dozens of round trips.
In the contact center, complaints about choppy calls pile up at peak times, when backups and updates fill the same line and queues form.
After the switch to softphones, telephony works flawlessly in the office, but variably in the home office, because Wi-Fi and consumer connections generate jitter.

Latency vs. bandwidth: the difference

Bandwidth describes how much data fits through a connection per second. Latency describes how quickly a single packet arrives. A vivid image: a wider highway carries more vehicles at the same time, but does not make a single car any faster. For backups and large downloads, bandwidth is therefore what counts above all; for interactive applications, RTT and jitter count. A bandwidth upgrade fizzles out when the delay is the real problem. It makes sense to measure both quantities separately and to define target values per application.

Latency optimization at KAEMI

KAEMI treats latency and jitter as measurable quality targets rather than as a chance product of the line. Through Cloud Connectivity & SDN , we connect your sites to cloud platforms via private interconnects and direct peerings, so that traffic takes short and controlled paths. QoS concepts prioritize real-time traffic such as voice and video along the entire route, and continuous monitoring makes the network values visible. For connections to Asia, China Connectivity reduces detours and packet loss on a route that is clean in regulatory terms. Talk to us before you commission the next bandwidth expansion: often the greater leverage lies in the path.

Frequently asked questions about Latency and Jitter

What is the difference between latency and ping?

Latency refers to the travel time of a packet in one direction. Ping is a measurement tool that determines the round trip time, that is the outbound and return path together. In practice the two terms are often equated. For application analysis, the RTT is usually the more relevant value, because protocols such as TCP wait for responses from the other side.

Which values are considered acceptable for VoIP and video conferences?

As a guideline, a delay of under 150 ms per direction applies so that conversations feel natural. Jitter should stay permanently below about 30 ms; packet loss in the low single-digit percentage range is already audibly disruptive. More important than individual peak values is stability over the day, because buffers compensate for constant delay better than for variation.

Does more bandwidth help against high latency?

In most cases, no. Bandwidth increases transport capacity, but shortens neither distance nor the number of hops. Only when a line is permanently overloaded and queues form does an upgrade also improve latency and jitter. Otherwise, direct paths with good peering and prioritization via QoS have a much stronger effect on response time.

How can jitter be reduced in the enterprise network?

Effective measures are prioritization via QoS and adequately dimensioned uplinks, supplemented by stable paths with as few intermediate stations as possible. Private connections to cloud platforms bypass congested internet sections. In the LAN, wired connections help for telephony and conference rooms. Added to this is monitoring that measures jitter continuously, so that changes in the network are noticed early.

Why is latency to China often so high?

Traffic between Europe and China covers large distances and passes through heavily utilized handover points as well as additional control instances at the network borders. Routes change frequently and are rarely optimized for business-critical traffic. Dedicated connections with defined handover points create predictable values. For this, KAEMI relies on legal and vetted connections instead of risky shortcuts.