130 likes | 250 Views
Signaling Compression: Overview and Questions. Carsten Bormann cabo@tzi.org based on slides from: Hans Hannu Hans.Hannu@epl.ericsson.se Jan Christoffersson Jan.Christoffersson@epl.ericsson.se Mats Nordberg Mats.Nordberg@epl.ericsson.se. Why?.
E N D
Signaling Compression:Overview and Questions Carsten Bormann cabo@tzi.org based on slides from: Hans Hannu Hans.Hannu@epl.ericsson.se Jan Christoffersson Jan.Christoffersson@epl.ericsson.se Mats Nordberg Mats.Nordberg@epl.ericsson.se
Why? • Minimize connection setup delay in complex 3GPP SIP interactions • Minimize bandwidth stealing for in-call usage of SIP • The point is not saving raw bandwidth (although it does help the network!)
What are the messages to be compressed? • SIP: • Largely a lock-step protocol • Essentially RFC822 (Text) • Can carry MIME payload • SDP: • v=2 m=audio etc. (Text) • Other MIME payloads are possible (SDPng!) • Either could be encrypted, possibly partially • RTSP (for streaming), also carrying SDP • DNS, RSVP, … ???
Why not IPCOMP (RFC2393)? • Yes, why not? • IPCOMP requires IPCA – need setup protocol (IKE?) • IPCOMP does not exploit inter-packet redundancies • Implementation issue: IPCOMP goes right into IP stack • May still be good enough, though: • Preloaded dictionary with SIP terms + LZSS 2.7:1 • Can use “manual configuration” using SRV • Or we could just steal IPCOMP’s framework?
Contributions • ROGER, draft-hannu-rohc-roger-01.txt • SCRIBE, draft-liu-rohc-scribe-01.txt • UDPcomp, draft-rosenberg-rohc-sip-udpcomp-00.txt • EPIC, draft-price-rohc-signaling-epic-00.txt • TCCB, draft-ziyad-rohc-tccb-00.txt
Signaling Compression: Components • 1) The protocol • Message handling, • E.g. Verification of correct decompression • E.g. Usage of previous messages in the compression process • E.g. Context state handling (dictionary/codebook handling),excluding algorithm-specific aspects • 2) The actual Compression Algorithm • What to save in the dictionaries/codebooks etc.. • Compressed message representation • E.g. Lempel-Ziv based representations Movable boundary
End to end (above the IP level) or per-link (below IP)? • Reordering • Link often can exclude reordering, e2e can’t • Negotiation issue • Link has PPP etc., how to negotiate e2e? • SRV-records/URL/…? • Piggy-back on security negotiation? • Encapsulation issues • How to match forward and backward direction? • What is most efficient?
Profiles • Could combine • One protocol with • One or more compression algorithms(plugged in like ROHC profiles) • future extensibility! Movable boundary
Focus of the contributions Larger X - more focus
Contribution specifics • Requires protocol knowledge to perform the actual compression • TCCB, EPIC (but falls back to ~ LZ) • Requires protocol knowledge for message handling • UDPcomp (uses application layer acks) • Can handle message reordering between compressor and decompressor • ROGER, SCRIBE, UDPcomp, (EPIC) • Well developed acknowledgement mechanism • ROGER, SCRIBE
Other metrics of the contributions *) Not counting performance data
Questions to the WG • Should it be done? • 3GPP wants it by R’5 (Dec 2001) • Should it be done here? • Will not be done in SIP WG any time soon! • Creation of new WG may take too long • Should not prejudice technical decision by WG choice • Discuss this in the light of the known solution space!
Questions to the presenters • What is the specific point of your proposal? • Is it protocol or algorithm or both? • How do its characteristics influence the WG decision? • How would it work with the other proposals? • Possible protocol/algorithm combinations?