This question and answer file contains APPN-related questions from APPN vendors and customers, and the answers provided by APPN architects and product designers, where appropriate. These Q&As deal with the following subjects, in that order, over the period Dec. '93 - Aug. '94. Subject: How does a router recognize APPN? Subject: Peripheral Border Node and Non-Adjacent APPN networks. Subject: APPN & HPR support for switched TGs Subject: IDBLK Numbers Subject: IDBLK number registration Subject: ANR performance in HPR products Subject: As400 as a Server Subject: APPN Base & Option Sets Subject: LLC & APPC Subject: APPN Source Code Subject: APPN Support Subject: APPN Conformance Testing Subject: APPN/LU6.2 over Ethernet Subject: Information on APPN Subject: APPN and switched networks ========================================================================= Subject: How does a router recognize APPN? Date: 12/13/93 Owner: Bruce Ferrero Compuserve?N Aetna Life & Casualty Question: Consider a multi-protocol router that must deal with IPX, pre-APPN SNA, IP, etc. Now it implements APPN network node routing as well. When a packet comes in, how does the router decide that the packet is APPN? And how solid is this algorithm? Is there any chance of confusion with other types of traffic, such as 3270 LU 2 traffic which may be coming in on Token Ring as 802.5? ANSWER: No fear. Unless the router has a really odd bug, it would never confuse the two types of traffic. Once it gets out into the network, the APPN traffic would have SNA headers with the appropriate session IDs. The other traffic would be encapsulated in TCP, using all of the TCP routing. This is not very different than what our APPN-capable controllers have been doing with SNA traffic already. Security(Public/Private): Public Architect: Steve Joyce ========================================================================= Subject: Peripheral Border Node and Non-Adjacent APPN networks. Date: 1/12/94 Owner: Dave Olson Compuserve?N Skill Dynamics Question: I. Without extended border node support, is it possible to establish an LU-LU session between non-adjacent networks? NETA NETB NETC | | AS/400 ------|----AS/400--------AS/400-----|------AS/400 Non-Native NN | Native NN Native NN | Non-Native NN | | LU-LU (Possible ?) <------------------------------------------------> My previous understanding had been that only one network boundary could be traversed. However, I recently saw the rule stated as: "The non-native network node partner for the peripheral boundary must be in the same subnet as the CP(OLU) or CP(DLU)." If that is the only restriction, then it would seem that the above configuration would allow an LU-LU session between non-adjacent networks. ANSWER: No, this will not work. Peripheral Border Node support supports a maximum of one network hop. You could establish a session between two resources in NETA and NETB, or in NETB and NETC, but you cannot establish a session between NETA and NETC unless a direct connection exists. When Extended Border Node support becomes available (VTAM now supports Extended Border Node in its V4R2 release), you would need to replace the two AS/400 NNs in NETB with Extended Border Nodes. Since Extended Border Nodes support multiple network hops, you would then be able to establish a session from NETA to NETC through intermediate network NETB. Security(Public/Private): Public VTAM Design: Roy Brabson Comments: ========================================================================= Subject: APPN & HPR support for switched TGs Date: 03/08/94 Owner: Adam Czesnick Compuserve? N Message#: Question: a) APPN architecture supports a single line TG between NNs.... Does the architecture provide for the support of switched TGs, which can be dialed up at times of heavy traffic load? b) Will HPR (which support multilink TGs) allow for dynamic addition of links to a transmission group? Answer: a) Yes, APPN does provide support for switched TGs. They are usually defined as limited resources, meaning that they are deactivated as soon as no sessions are no longer active over them, and then reactivated again when they need to be used. For more details on how this affects APPN configurations and APPN routing, see the "Switched Link Connections" section on p. 9-11 in the SNA APPN Architecture Reference manual (SC30-3422-03). b) Neither APPN nor HPR provides support for multilink TGs. Security(Public/Private): public Architect: Tim Hefel Comments: ------------------------------- Referenced Note --------------------------- Subject: IDBLK Numbers Date: 03/11/94 Owner: William Der Compuserve? N Message#: Question: This question concerns the specification of the IDBLK parameter used in definition of link stations for an APPN network node. The IDBLK parameter is used during the XID-3 negotiation proceeding exchanges to determine a primary or secondary role for a type negotiable link station. IDBLK is actually carried in bytes 2 - 5, bits 0 - 11 in the XID-3 frame. Typically, IDBLK represents an IBM product ID. For instance, for OS/2 CM/2 network nodes, IDBLK defaults to 05D. For AS400 network nodes, IDBLK defaults to 056. Given the openness of APPN with multiple vendors participating in the implementation of APPN products, there seems to be a need to assign unique vendor/product specific IDBLK numbers to non-IBM vendors. I'd like to know if IBM already does this, and if so, how I might go about getting IDBLK numbers assigned for my products. Answer: I am the registrar for the block ID numbers. I will be happy to assign one or more numbers for use by your products. I just need a brief description to associate with each number in the registry, and your name and address (company, city, and state is fine) as the requester. Security(Public/Private): Public Architect: Mike Allen (moallen@vnet.ibm.com) Comments: ======================================================================== Subject: IDBLK number registration Date: 03/11/94 Owner: Bill Der Compuserve? N (Cabletron) Message#: Question: Mike, Thanks for your response regarding IDBLK numbers. Below is the information you requested in order to assign us an IDBLK number. Description: Cabletron APPN Router The registration will be under: Cabletron Systems, Inc. 35 Industrial Way P.O. Box 5005 Rochester, NH 03867-0505 Would it be convenient at this time to assign to us a block of 10 IDBLK numbers to identify Cabletron APPN products running on different Cabletron boards or product lines in the future, or should we continue to request them on an as needed basis when more detailed descriptions are available? Thanks again, Mike. ANSWER: Bill, I have registered block ID number X'E08' for Cabletron's APPN router. We don't reserved blocks of numbers for future use, so you will need to send us a specific request if you need another number in the future. However, you don't necessarily need a different block number to represent every APPN product you sell if you can guarantee uniqueness in the IDNUM value across all the copies of all your products. If managing the IDNUM values within a block number is problem, I will be glad to assign additional block numbers to represent each product. Security(Public/Private): Public Architect: Mike Allen (moallen@vnet.ibm.com) Comments: ========================================================================= Subject: ANR performance in HPR products Date: 03/14/94 Owner: Adam Czesnik Compuserve? N IBM Canada Ltd. Message#: Question: In HPR presentations and other doc, we quote ANR performance as 3X and up over ISR performance. I'd like to clarify this so I deliver a message consistent with tests/simulations. 1. What are the test variables (in general terms) used for derivation of these numbers? Link speeds? What are they (56kb and up?) Is the node to node connection a WAN connection? Platform which HPR node is implemented on?(AS/400, VTAM/NCP etc) 2. Does "performance" comparison mean throughput over an RTP connection vs throughput over an LU-LU session in an ISR implementation, or does it have another meaning. ANSWER: The comparison was purely based on the pathlength that it takes to do ISR vs. ANR. We counted the ISR pathlength on AS/400 and 6611 router and estimated how long it takes to do ANR across these plat- forms. The numbers estimated for ANR were rather conservative. That's how the 3X gain in performance was obtained. The pathlength to do ISR or ANR is not dependent on packet length. Hope this helps Security(Public/Private): Public Architect: Lap Huynh Comments: =============================================================================== SUBJECT: As400 as a Server DATE: 3/30/94 OWNER: Ian Addison - EMI Music QUESTION: Hi this is a call of desperation I am the IT Director for EMI Music in the UK . We run As400's (big ones), MVS mainframes multiple Lans and other things I have a large IT budget. I am keen to use AS400 as a server. But IBM just have not got their act together. They say all the right words but where is the substance. I need to contact some guys that can make it happen. Local sales try but to no availe (no crtisism is being leveled at them) We have the people. We want to bench mark major appliactions. We need access to experienced people who have done it with gigabyte data files linked to MIS/EIS applications. That understand what has to be done and can help with our benchmarks.(we are in the shortterm confined to powerbuilder on the client). Yes we have the latest Os400 and the odbc drivers ready to go. I know about some of the work in the states but it aint hittin me. I know about the last road show but it was poor and badly structured. Who can drive this. ANSWER: "What makes a good server" PMJI - I have this great little paper from IBM called "Application Reference Designs for Distributed Servers". It describes many of the functions that a server may have to perform. These are described from a platform-neutral point of view. While I am not an AS/400 expert, I see no reason why it could not offer these functions. I do a fair bit of training and a little consulting. I meet quite a few people. You may have problems finding odbc/powerbuilder experts. I have not met many. If you have to go the SQL route, have you thought about VisualAge instead of PowerBuilder? From what I've seen it has some interesting features. Again I have not met many experienced VisualAge experts. I think I am going to be in London next month. I would be very interested in having a chance to talk about your experiences with client/server and the AS/400. Do you know if IBM offers OSF/DCE service's under OS/400? Q: Jerry, where can I get a hold of this paper? Sounds useful. Well.. that's the problem. I got my copy from my IBM Consultanting contact. There is no identifcation (even a publication number) on the paper. When I asked my conact where they found it they did not know! I have been searching for the source ever since. However, if you can send me a fax number by e-mail I would be happy to send you a copy next week when I get home. RESPONDENT: Jerry Golick - 71175.1011@compuserve.com ANSWER: The AS/400 will be supported as a server in the future May 3rd [1994] I think (Hint, Hint) RESPONDENT: Mitch Kleinman COMMENTS: =============================================================================== SUBJECT: APPN Base & Option Sets DATE: 4/11/94 OWNER: Phil Rutherford QUESTION: I have just finished doing some testing of VTAM (V4R1) APPN support together with CM/2 running on a number of LAN attached NN and ENs. VTAM and NCP are defined as a composite node. The 3745 is token-ring attached to the same LAN as the OS/2 systems. VTAM is defined as a CDS. [Central Directory Server] One of the CM/2 systems is defined as a NNS for several CM/2 ENs on the LAN. There is a CP-to-CP session between the NNS and VTAM (NN). When I bring up the ENs they register their resources with their NNS (I see this in the trace) but the NNS does not register these resources with the VTAM CDS, (I can not see any evidence of this in the trace). According to the APPN Architecture manual (Appendix A) CDC (Central Directory Client) is a Base function (065) and yet it does not seem to be implemented in a release of CM/2 from November 93 (V1.1). When can we expect to see this base function in CM/2? Is there any ducument which documents which APPN Base functions and Option Sets are implemented in which products? For example there is A WSC manual which documents which APPC option sets are implemented in which products. Another area where I received different results during testing to those I would have expected was in the area of NNS failure/takeover/give back. I wanted to define a primary NNS on the LAN and use the VTAM NN as a secondary NNS. If I defined CP-to-CP connection between an EN and its NNS and the same EN and VTAM it was imposible to control which NN became the NNS. In the event of a failure of the PNSS then in order to take over control from another NNS it was necessary to break all the sessions. What is needed is a an ACQUIRE/RELEASE type function which is non-disruptive (a bit like SSCP takeover/give back). Are there any plans to deliver function like this in the VTAM and CM/2 world? Base function 1015 talk about functions along these lines. When can we expect to see this type of function implemented? Thanks for your help - Phil ANSWER: Here is the reply from my architect friend... <> Editor's note: the new version is now available, and can be ordered using the product # listed aboveŁ RESPONDENT: Zeke Crater ANSWER: The following function sets are derived from figures in Appendix A of the SNA APPN Architecture Reference. APPN CM/2 1.1 END NODE FUNCTION SETS SUPPORTED ----------------------------------------------- 008 - Multiple TGs 023 - BIND Segmenting and Reassembly 024 - Adaptive BIND Pacing (Dependent) 038 - Limited Resource 041 - Session Segmenting/Reassembly 081 - Class-of_Service Manager 150 - MDS Common Base 151 - MDS End Node Support 160 - MS_CAPS Base En Support 161 - MS_CAPS Have Backup or Implicit FP 162 - MS_CAPS Be Sphere of Control (SOC) EN 170 - EP Alert Base Subset 171 - Problem Diagnosis Data in Alert 177 - LAN Alert 178 - SDLC/LAN LLC Alert 181 - X.25 Alert 182 - Held Alert for CPMS 1002 - Adjacent Link Station Name 1007 - Parallel TGs 1011 - Multiple Local Independent LUs 1012 - CP=LU 1200 - Tree and TG Caching 1520 - Common Operations Services APPN CM/2 1.1 NETWORK NODE FUNCTION SETS SUPPORTED --------------------------------------------------- 024 - Adaptive BIND Pacing (Dependent) 038 - Limited Resource 065 - Central Directory Client 105 - Intermediate Session Routing for Dependent LUs 150 - MDS Common Base 152 - MDS NN Support 161 - MS_CAPS Have Backup or Implicit FP 163 - MS_CAPS Base NN Support 164 - MS_CAPS Have a Subarea FP 170 - EP Alert Base Subset 171 - Problem Diagnosis Data in Alert 177 - LAN Alert 178 - SDLC/LAN LLC Alert 181 - X.25 Alert 182 - Held Alert for CPMS 1002 - Adjacent Link Station Name 1007 - Parallel TGs 1011 - Multiple Local Independent LUs 1012 - CP=LU 1101 - Preload Directory Cache 1104 - TDB-Based Nonverify 1200 - Tree and TG Caching 1520 - Common Operations Services RESPONDENT: Zeke Crater COMMENTS: =============================================================================== SUBJECT: LLC and APPC DATE: 4/13/94 OWNER: Andrew G. Labrot QUESTION: I have a situation where a potential customer is running a pure TCP/IP network and has only been able to access their mainframes via channel attched RISC6000 boxes. They are now interested in using APPC to gain access to their CICS regions. Can APPC run natively over TCP/IP? Does APPN support have to be present? Do they need LLC support? Can LU6.2 be cleanly and natively encapsulated into TCP/IP without SNA anywhere? They also use various combinations of Network Management , OS and workstations. ( I beleive they are from the school of "I don't have any bad hardware/software..I have it all") All I have read sheds no light on these fundamental questions. I have read my Red Books, CD ROMS and other literature, yet I can't seem to find a definitive answer anywhere. If you could take a minute and respond, I would really appreciatte it. ANSWER: Today, APPC will not run over TCP/IP from a RISC6000 to an MVS host. APPC needs some APPC platform on the machine (like SNA Server/6000). If the RISC6000s were LAN attached to the host, SNA Server/6000 would allow you to talk APPC over the LAN to the host. IBM's AnyNet products which allow APPC to flow over TCP/IP and TCP to flow over APPC do just what you want. IBM has issued a statement of direction for an AnyNet/6000 product but it has not been announced yet. RESPONDENT: Zeke Crater COMMENTS: =============================================================================== SUBJECT: APPN Source Code DATE: 4/26/94 OWNER: Anders - Axis Communications QUESTION: I'd like up-to-date information about licensing APPN source code. Where do I go? Is there a licensing/royalty fee for each produced unit? ANSWER: Download APPNOW.ZIP from Library 4, Technical Papers. It should answer all your questions. [This library is located on the APPC forum on CompuServe. Type GO APPC or GO APPN to access it] RESPONDENT: Zeke Crater COMMENTS: =============================================================================== SUBJECT: APPN Support DATE: 5/27/94 OWNER: Dave Savy - Intersolv France QUESTION: Can somebody reply to the following questions(may be they are silly and irrelevent). Thanks in advance 1. Can a NETWARE FOR SAA Gateway provide APPN capablities ? 2. If so does it provide LU 6.2 support ? 3. And can the gateway route sessions upstream to a MAINFRAME (IBM 3090) and downstream to the LAN BASED (ethernet) workstations (either DOS or OS/2) using NS/DOS, CM/2 and/or RUMBA TOOLS for APPC. I'd appreciate your replies ANSWER: A Netware for SAA Gateway runs as a LEN Node. So it can participate in an APPN network but it cannot provide APPN routing. The Gateway can route sessions from the Netware clients to the mainframe and back. However, since it is a LEN node, you cannot route through the Gateway to other End Nodes or Network Nodes running CM/2, NS/DOS or the RUMBA Tools for APPC. The Netware for SAA Gateway does provide APPC (LU 6.2) support. RESPONDENT: Jill Bodine COMMENTS: =============================================================================== SUBJECT: APPN Conformance Testing DATE: 6/02/94 OWNER: Anthony Johnston QUESTION: I understand that a number of companies (Wellfleet etc) have actualy licensed APPN code from a UK based company (DCL??), rather than from IBM. This suggests that perhaps this code was independantly written by DCA (presumably from IBM's APPN design documents). If this is the case, given that each "standard" (especially it seems in Networking) is open to Interpretation by the implementor, what mechanisms are in place to ensure that each "implementation" of APPN support can actualy co-exist and fully pariticpate in an APPN network? It would be a great pity if these problems of interworking were to beset APPN as they did in the early implelentations of X.25 in the late 70's/early 80's. ANSWER: IBM continuously works with other vendors who implement APPN to insure interoperability with their platforms. In fact, IBM is sponsoring an APPN Connect-a-thon this June 21-23 to conformance test and test product interoperability of APPN implementations. 9 companies, including DCL, are taking part in this Connect-a-thon. We are also jointly constructing an APPN test specification to provide companies a way to "self-certify" their APPN nodes, through the APPN Implementors Workshop (AIW). An Internet mailing list has been created to allow any interested developers/testers an opportunity to participate in this work, or just to learn about what is being done to insure interoperability between APPN products. If you or anyone else is interested in joining this mailing list, send a note to: aiw-validation-request@ibmstandards.cary.ibm.com. Also, if you'd like to examine the APPN test specification, it is available via anonymous FTP to the IBMstandards repository. Simply FTP to ibmstandards.cary.ibm.com. The test file is located under directory pub/standards/aiw/validation/doc, and is called aiwtspec. It is available in PostScript form, as well as zipped PostScript form. RESPONDENT: Tim Hefel RESPONSE FROM ANTHONY: Thanks for the information. I'm happier that IBM is taking (more than) an interest in the future of what I believe will a major networking technology in the next few years. I realise that in many ways IBM is in a "no-win situation", if IBM exert a vice like grip over APPN directions then you will be criticised, and if you "let" different implementations of APPN flourish then again you will be criticised..... I've been pilloried by some of my collegues for my "faith" in SNA whilst they are flying the "OSI" flag of interoperability, but I see APPN and the Networking Blueprint/MPTN announcements as a major step forward for IBM and its customers who have a real business requirement to communicate to other platforms/beyond the "blue barrier"! COMMENTS: =============================================================================== SUBJECT: APPN/LU6.2 over Ethernet DATE: 6/23/94 OWNER: Barry Edelman QUESTION: Question: Can a UNIX host supporting PU2.1 communicate with VTAM via Ethernet? I'm interested in accessing my mainframe (ES/9000, MVS) from UNIX server using LU6.2 sessions. Traditionally, I'd attach the UNIX system to the token ring network and go straight to VTAM via the 3745 TIC. For several reasons, this UNIX system can only attach to a Ethernet segemnt. The Data Center has a 3172-001 for testing TCP/IP (we're new to this). Suggestions? BTW, it was really a plesant surprise to see a forum on CompuServe dedicated to this subject! [this question originally posted on the APPC CompuServe forum, under APPCARCH section.] ANSWER: I can see no reason why this shold not work. We use a 3172 clone to connect a large DEC network to an SNA (PU2.0) network. You've probably answered the question yourself when you said that you would usually attach the UNIX box via Token Ring. VTAM doesn't really understand (or care about!) DLC protocols like SDLC/Token Ring Ethernet etc. VTAM only understands the IBM channel protocol - via CCW's etc. (with the exception of some mainframes with integrated communications adapters + a special VTAM(E)). A box such as the 3172 or 37XX is required to provide the DLC to channel protocol conversion. So if your UNIX box can handle the PU2.1 stuff then you should be Ok. Hope this makes sense! RESPONDENT: Anthony Johnston COMMENTS: =============================================================================== SUBJECT: Information on APPN DATE: 6/26/94 OWNER: Chris Setz QUESTION: How does APPN work? If you can point me to online sources, I would be grateful. I have 15 years in the industry, and an LAN/WAN centred. Thanks in anticipation, ANSWER: APPN works well :-). I think it is great. Take a look at APPNE.ZIP and EN21.ZIP in library 11, APPN Architects Only, and APPNOV.ZIP, APPNOW.ZIP, and APPNST.ZIP in library 4, Technical Papers. These are all good sources of info on APPN and how it works. APPN is the way to go! Let me know if you have any further questions. APPN: the smartest distance between two points! Editor's note: other good sources for info. regarding APPN are 1) APPN Architecture and Product Implementations Tutorial - 6/94 (order # GG24-3669-02) 2) Best of APPC, APPN, and CPI-C (order # SK2T-2013-03) 3) APPN Architecture Reference (order # SC30-3422-03). These books can be ordered through IBM MechanicsburgŁ RESPONDENT: Zeke Crater COMMENTS: ========================================================================= Subject: APPN and switched networks Date: 8/17/94 Owner: Alan Barker-Benfield Compuserve?N Question: I am working in a pure CM/2 world and there are ONLY switched (ISDN) links between the NNs. Autoactivation works. What is not working yet however is the 'finding' of parner LU's on EN nodes that can only be reached across the switched connection. The link activates automatically and then I receive sense 8004000. I am trying to understand if this is an architectural limitation or a user definition error. Can you help ? ANSWER: RG: This is a definition error. You should get APPCINF or GETSENSE package for an explanation. JF:What you need is LU registration function of APPN. With this function, the ENs' could register their applications and then the network node would know on which EN to find the application. BUT, as APPN is currently implemented, I BELIEVE that registered LUs are "forgotten" when the connection to the EN is dropped, so its a catch-22 ... What we need is to remember the resource locations when we drop the CP-CP session (connection) to the EN. Then when we get a request for the application, we would reactivate the CP-CP session and establish the session to the EN's appl. But I'm just dreaming ... CM:The simple answer is that, if you've just got switched links, you've got a configuration where every, in your view, End Node, is a LEN Node and every Network Node connects to every other adjacent node as if it were a LEN node, whether it's configured as a LEN Node or a Network Node. This means all partners need to be defined, including in the Network Nodes which you want the session to traverse. Since you have no CP-CP sessions, you have no directory or topology data gathering support. What you need is a "cheat" solution such as the one used by X.25 NPSI which, for subarea links using virtual calls, or switched virtual circuits, SVCs, can set up calls as bursts of packets arrive but clears the call after a period of idleness. Consultants: Richard Gray (RG) APPC Market Enablement Jim Fletcher (JF) VTAM Design Chris Mason (CM) Comments: