Performance of Splitting LTE-WLAN Aggregation

Abstract

LTE-WLAN aggregation (LWA) combines the radio resources of LTE and WLAN to take advantage of Wi-Fi’s high availability and indoor coverage, which provides better usage of both WLAN and LTE. In LWA, a user data radio bearer (DRB) is split if its packets are delivered through both LTE and Wi-Fi. For a split DRB, it is important to determine the LTE-to-WLAN ratio (LWR) or the ratio of packets delivered by LTE over WLAN. Based on WLAN’s Received Signal Strength Indicator (RSSI), we propose a simple LWR selection rule. Using this rule, we describe a user plane implementation for LWA, where the adaptive LWA routing procedure can be easily implemented in the Radio Resource Management (RRM) and the Packet Data Convergence Protocol (PDCP) layer at the LTE eNB. In our implementation, the maximum LWA throughput is 99.14 percent of the optimal cases, and the packet loss rates are less than 0.0042 percent.

This is a preview of subscription content, log in to check access.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

References

  1. 1.

    3GPP TR 36.889 v13.0.0 (2015-06), Study on Licensed-Assisted Access to Unlicensed Spectrum

  2. 2.

    Sirotkin S (2017) LTE-Wireless Aggregation (LWA): Benefits and Deployment Considerations. Intel White Paper

  3. 3.

    Lin Y-B, Shih Y-J, Chao P-W (2018) Design and implementation of LTE RRM with switched LWA policies. IEEE Trans Vehic Tech 67(2):1053–1062

    Article  Google Scholar 

  4. 4.

    Mobile World Congress (2016) Barcelona

  5. 5.

    3GPP TS 36.300 v13.4.0 (2016-06) Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall description; stage 2

  6. 6.

    Tu Y-K et al (2017) Deployment of the first commercial LWA service. IEEE Wirel Commun 24(6):6–8

    Article  Google Scholar 

  7. 7.

    Industrial Technology Research Institute (2016) ITRI LTE-A Small Cell System and Smart Antenna Technology. http://wireless.itri.org.tw/technology/#LTE

  8. 8.

    Texas Instrument. TCI6638K2K Multicore DSP+ARM KeyStone II System-on-Chip. http://www.ti.com/product/tci6638k2k

  9. 9.

    Texas Instruments (2012). Keystone Architecture Packet Accelerator (PA) User Guide

  10. 10.

    Texas Instruments (2013) Keystone Architecture Security Accelerator (SA) User Guide

  11. 11.

    3GPP TS 29.281 v13.2.0 (2016-06) General Packet Radio System (GPRS) Tunnelling Protocol User Plane (GTPv1-U)

  12. 12.

    3GPP TS 36.323 v13.2.0 (2016-07) Evolved Universal Terrestrial Radio Access (E-UTRA); Packet Data Convergence Protocol (PDCP) specification

  13. 13.

    3GPP TS 36.322 v13.2.0 (2016-07) Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Link Control (RLC) specification

  14. 14.

    iPerf. https://iperf.fr

  15. 15.

    3GPP TS 36.331 v13.2.0 (2016-07) Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Resource Control (RRC); Protocol specification

Download references

Acknowledgements

This work was supported in part by Ministry of Science and Technology (MOST) 106N490, 106-2221-E- 009-049-MY2, 107-2221-E-009-039, ``Center for Open Intelligent Connectivity” of National Chiao Tung University and Ministry of Education, Taiwan, R.O.C.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Li-Chang Wang.

Appendix: Packet loss performance

Appendix: Packet loss performance

This Appendix describes how to compute the packet loss in LWA. We calculate the packet loss according to the PDCP Status Report obtained from the UE [12]. The parameters used to calculate the packet loss are described as follows.

  • The First Missing PDCP Sequence Number (FMS) is the sequence number (SN) of the first missed PDCP PDU. If no missed PDU is found, the FMS is set to the SN of the next PDU to be transmitted. We note that a missed PDU may not be lost. This PDU may be delivered out-of-order, and will be received later.

  • The Bitmap records the receiving status (i.e., 0 for missed and 1 for received) of PDUs with the SNs following the FMS. If there is no out-of-order PDUs, the Bitmap is empty. We use Fig. 8 as an example to illustrate the relationship between the FMS and the Bitmap, where the Bitmap size is 8 (in ITRI LWA, the Bitmap size is 16384). In Fig. 8a, PDUs 0-10 are sent to the UE and PDUs 2, 5 and 7 are missed. At this point, the FMS and the Bitmap of the PDCP Status Report are 2 and 11010111, respectively.

  • PDCP Status reporting periodicity ts specifies the period of PDCP Status Report sending [15], which is configured by the RRC/RRM. The period is set to 20 ms in our experiments.

  • PDCP reordering timer tr is configured by the RRC/ RRM. The UE’s PDCP starts this timer when it detects a missed PDU [15]. If the PDU indicated by the FMS is not received when tr expires, then it is considered lost. On the other hand, if the PDU indicated by the FMS is received before tr expires, then tr is stopped by the UE. In Fig. 8a, suppose that tr expires and PDU 2 is still not received by the UE. Then PDU 2 is considered lost, and the FMS pointer is moved to the next missed PDU specified in Bitmap, which is PDU 5 (Fig. 8b). When the FMS is updated, the received packets between the previous FMS = 2 and the next FMS = 5 (i.e., PDUs 3 and 4) are sent from the PDCP layer to the IP layer in the UE. On the other hand, if PDU 2 is received before tr expires, then tr is stopped. The FMS pointer is moved to PDU 5 (Fig. 8c), and the packets received before the next FMS (5) (i.e., PDUs 2, 3 and 4) are sent to the upper layer.

Fig. 8
figure8

The relationship between the FMS and the Bitmap: a FMS = 2; b FMS = 5 (PDU 2 is lost); c FMS = 5 (PDU 2 is received)

We conducted experiments to measure the throughput under different tr values, and use these results to select an appropriate tr period. The curves in Fig. 9 show that the maximum LWA throughput is reached when tr is 60 ms. Since tr and ts may not be synchronized (i.e., tr is triggered by the UE when a PDU is missed, while ts is triggered by the UE to periodically send PDCP Status Reports to the eNB), tr should not be shorter than 60 ms. However, a longer period of the PDCP reordering timer implies a larger PDCP buffer size in the receiving end. In ITRI LWA, we set tr to 80 ms to prevent from miscalculation of packet loss, while the buffer size is not too large.

Fig. 9
figure9

Effects of tr on the throughput

Through tr and ts, the packet loss is determined at the Routing subtask in the PDCP. When the eNB receives [tr/ts] PDCP Status Reports with the same FMS and the Bitmaps are not empty, the PDU indicated by the FMS is considered lost. Because tr and ts are set to 80 ms and 20 ms, respectively, we consider the PDU indicated by the FMS is lost after [tr/ts] = 4 PDCP Status Reports with the same FMS value are received. Therefore, the packet loss calculation procedure is proposed as follows.

Let Si = FMSi,Bi be the i th PDCP Status Report received by the eNB, where FMSi and Bi are the ith FMS and Bitmap, respectively. Suppose that there is no missed PDU when the UE sends Si to the eNB, then FMSi is the SN of the last received PDU plus one, and Bi is empty. Suppose that a PDU is missed before Si+ 1 is issued, then when the missed PDU is detected by the UE, tr is triggered. When the eNB receives Si+ 1 from the UE, it finds that Bi+ 1 is not empty, FMSi+ 1FMSi, and a counter C at the Routing subtask is set to 1. This counter is used to record the number of PDCP Status Reports received during a tr period. Then the eNB continues to receive Si+ 2, where one of the following three cases occurs.

Case 1.:

Bi+ 2 is empty. In this case, tr has been stopped by the UE before Si+ 2 is issued, and one of following two subcases occurs.

Case 1.1.:

If C ≥ [tr/ts], the PDU indicated by FMSi+ 1 is considered lost.

Case 1.2.:

If C < [tr/ts], the PDU indicated by FMSi+ 1 and the subsequent PDUs are received.

In either case, FMSi+ 2 is the SN of the next PDU to be received, and the PDUs with the SNs between FMSi+ 1 and FMSi+ 2 are sent from the PDCP layer to the IP layer in the UE.

Case 2.:

Bi+ 2 is not empty and FMSi+ 2 = FMSi+ 1. In this case, the PDU indicated by FMSi+ 1 has not been received. C is incremented by one.

Case 3.:

Bi+ 2 is not empty and FMSi+ 2FMSi+ 1. In this case, FMSi+ 2 is the SN of the next missed PDU, which implies that tr has expired and been retriggered before Si+ 2 is issued. Therefore, C is reset to 1 after one of the following two subcases occurs.

Case 3.1.:

If C ≥ [tr/ts], the PDU indicated by FMSi+ 1 is considered lost (e.g., Fig. 8b).

Case 3.2.:

If C < [tr/ts], the PDU indicated by FMSi+ 1 has been received (e.g., Fig. 8c).

We conducted experiments to calculate the WLAN packet loss rates and the results are shown in Fig. 4.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lin, Y., Tseng, H., Wang, L. et al. Performance of Splitting LTE-WLAN Aggregation. Mobile Netw Appl 24, 1587–1595 (2019). https://doi.org/10.1007/s11036-018-1179-8

Download citation

Keywords

  • Data Radio Bearer (DRB)
  • LTE-WLAN aggregation (LWA)
  • LTE-to-WLAN ratio (LWR)
  • Received Signal Strength Indicator (RSSI)
  • Split DRB