Definition
Differentiated Service (Differentiated Service) provides different levels of service quality (SLA) based on business type, user level, payment level, etc. 5G network slicing is a typical example of differentiated service, where the same physical network can provide virtual networks with different latency, bandwidth, and reliability requirements such as eMBB, mMTC, and uRLLC. It works together with QoS, Bandwidth Management, and Priority QoS to achieve fine-grained resource allocation.
Implementation Methods
- Network Slicing: Under the 5G SBA architecture, different S-NSSAI correspond to different slices, isolating resources and policies
- QCI/5QI Mapping: Bearer-level QoS, mapping services to different QCI (4G)/5QI (5G), with differentiated scheduling and packet loss policies
- Dedicated APN/DNN: Private network access , assigning enterprises independent access points and policies
- Bandwidth Reservation and GBR: Configure GBR to guarantee bit rate for critical services
IoT Scenarios
IoT devices come in various types: video monitoring requires high bandwidth and low latency; sensor reporting can tolerate delays and uses low priority; vehicular networks require uRLLC. Differentiated services allocate different resources to different services through QCI/5QI, slicing, and private networks.
Typical Use Cases
Enterprise private networks, vehicular networks , industrial IoT , video monitoring, etc. SimtoLink can configure differentiated QoS and private network strategies based on customer business types.
Differentiated service is SimtoLink's core capability to provide customized connectivity for enterprise customers. Through dedicated APNs, QCI/5QI mapping, bandwidth reservation, etc., SimtoLink can allocate differentiated resources to IoT devices of different business types. The SimtoLink cloud platform supports device-based and service-based QoS configuration to meet enterprises' needs for high-priority guarantees for critical devices.
A differentiated-services code point is rewritten as a bearer class inside the cellular network
An enterprise network often marks voice, video, and default traffic with a differentiated-services code point in the header. Once the packet enters the cellular core, what the scheduler looks at is often no longer that code point. It is the quality-of-service class on the bearer. If the gateway does not map the code point to the matching class, the mark is as good as unwritten on the air interface. The device can set the code point very high, and the base station still forwards on a best-effort default bearer. To see whether the mapping works, capture on both sides of the gateway. If the entrance shows the code point and the exit bearer identity did not change, the mapping table is missing that row.
The reverse direction matters too. If a cellular class is not written back as a code point when it reaches the enterprise firewall, the LAN on the second half treats every packet as the same class. Differentiation needs a matching action on every hop. A mark on the first hop does not make later queues classify on their own.
Two flows on the same device can land in different classes
The unit of differentiation can be a bearer or a service flow. It does not have to be the whole card. An alarm report and a log upload can leave the same module and, if they match different packet filters, one can take a low-delay class and the other the default class. Filters usually look at address, port, or protocol. If the port is wrong, both flows fall into the default bearer, and alarm delay jitters like the log. Acceptance should send both kinds of packet on the real ports and read the bearer identity of each. Do not send only one probe packet.
When filters overlap, the first matching rule wins. Put “all uplink” in front of “the alarm port,” and the alarm rule is never reached. Order the table with the more specific condition first and the wider condition after it, and after a change send one alarm packet to confirm it did not fall into the wide rule.
When a slice and a bearer priority both exist, decide which one takes effect first
A slice sets aside a pool of resources. Bearer priority decides who is scheduled first inside that pool. If the device entered the wrong slice, a high bearer priority only goes first inside the wrong pool. It never touches the part prepared for low delay. The other way around, a correct slice with every bearer on the default class can still leave a critical flow behind heavy traffic in the same slice. Both settings have to be checked. One cannot stand in for the other.
What SimtoLink can align with the enterprise is the access point, the bearer class, and subscription policy, the differences that can sit between the card and the network. If the application’s own queue priority is not mapped onto a bearer, the network side cannot see it. The platform should be able to show the current bearer class per device. If the class stays on the default for a long time and the service needs low delay, change the filter and the subscription before only raising a thread priority in device software.
A mark on the default bearer does not grow into a dedicated bearer by itself
If a device marks packets as high priority on the default bearer, that only affects how packets queue inside the bearer that already exists. It does not create a dedicated guaranteed-bit-rate bearer out of nothing. A dedicated bearer has to be set up by the network from the subscription and the filters. If the mark is present and the bearer identity did not change, the scheduler still assigns resources at the default class. Acceptance cannot look only at the mark the device sent. It has to see whether the network sent a new bearer setup. Without that signaling, every later priority is still on the original bearer.
When bearer setup is rejected, the service should have a clear fallback: keep using the default bearer and tell the application, rather than retrying setup without limit. Retries add signaling during congestion. What SimtoLink can help check is whether the subscription contains the class, and whether the filters were delivered onto the network. Application thread priority and process weight never reach this bearer. If the platform shows the default class for a long time and the service needs a fixed delay, finish bearer setup first, then talk about marks and queues. If the order is reversed, a finely tuned queue is only sorting inside best effort.