diff options
Diffstat (limited to '2019/osmodevcon2019-osmocom_and_lte/osmodevcon2019-osmocom_and_lte.adoc')
| -rw-r--r-- | 2019/osmodevcon2019-osmocom_and_lte/osmodevcon2019-osmocom_and_lte.adoc | 92 |
1 files changed, 92 insertions, 0 deletions
diff --git a/2019/osmodevcon2019-osmocom_and_lte/osmodevcon2019-osmocom_and_lte.adoc b/2019/osmodevcon2019-osmocom_and_lte/osmodevcon2019-osmocom_and_lte.adoc new file mode 100644 index 0000000..bd71b2f --- /dev/null +++ b/2019/osmodevcon2019-osmocom_and_lte/osmodevcon2019-osmocom_and_lte.adoc @@ -0,0 +1,92 @@ +Osmocom and LTE +=============== +:author: Harald Welte <laforge@gnumonks.org> +:copyright: 2019 by Harald Welte (License: CC-BY-SA) +:backend: slidy +:max-width: 45em + + +== Osmocom and LTE + +* most of Osmocom CNI is written by paid developers these days +* developments mostly determined by customer requirements +* sysmocom isn't facing any significnat LTE requirements for years +* Result: No _native_ LTE support in any Osmocom components +* However: we do have small LTE related bits +** SI2quater for broadcasting eUTRAN neighbors over GSM +** SGs interface for 2G/3G side of CSFB and for SMS-over-SGs + +== Irrespective of Osmocom... + +What do we need to run a combined 2G/4G network? + +1. shared subscriber database (HLR + HSS) +2. advertisement of respective neighbor cells +3. shared PS tunnel endpoint (GGSN + P-GW) + +optionally: + +* SGs interface in MME and MSC for CSGB + SMS-over-SGs +* Inter-RAT PS handover +* SRVCC (Single Radio Voice Call Continuity) + + +== Shared Subscriber Database + +* Required mainly for Authentication +* poor man's hack: Simply install same IMSI/Key material in both HLR and HSS +** will result in auth re-sync on every authentication due to non-synchronized SEQ/SQN of UMTS AKA +* proper solution: +** Single HLR with both GSUP and DIAMETER, or +** DIAMETER-to-GSUP translator (to make LTE use OsmoHLR), or +** GSUP-to-DIAMETER translator (to make OsmoMSC/SGSN use HSS) + +== Advertisement of neighbor cells + +* GSM must advertise eUTRAN neighbors +** we have SI2quater support in osmo-bts +** TEMS phones show proper/valid decode +* LTE must advertise GSM neighbors +** no "OML" for S1-AP +** configuration mechanism is vendor-proprietary + +== Shared PS tunnel endpoint + +* if the MS/eUE moves from GPRS to LTE, it wants to keep PDP/PDN contexts +* solution: implementing P-GW and GGSN functionality in one element +** ergw might be able to do that (untested, complex) +** OsmoGGSN doesn't speak GTPv2C (required for P-GW) +** nextepc P-GW doesn't speak GTPv1C (required for GGSN) + +Solution: + +* play with ergw, or +* adopt nextepc GTPv2C code generator to use msgb/libosmo* and add GTPv2C to OsmoGGSN + +== SGs interface + +* Osmocom already implements full SGs interface +* Interaction with MME on LTE side needs only interop testing + + +== Inter-RAT PS handover + +* inter-RAT HO offers seamless migration between GERAN + eUTRAN +* uses S3 interface between MME and SGSN (GTPv2C based) +* uses S4 interface between S-GW and SGSN (GTPv2C based) +* tbh, I don't know the details here. +* any help appreciated +* https://www.tech-invite.com/3m23/toc/tinv-3gpp-23-401_zh.html +* probably a long way for Osmocom to support it, given that we don't really do any inter-SGSN HO even inside a single RAT. + +== Conclusions + +The logical steps to improve our LTE interworking are: + +1. split HLR/HSS and split GGSN/P-GW (today) +2. GSUP-to-DIAMETER translator for shared HLR/HSS +3. shared GGSN/P-GW for + +== EOF + +End of File |