IBM CIP Protocols

A. 16 MB is recommended for CSNA and Offload. Individually, 8 MB is sufficient. The default DRAM is 32 MB on CIP2 modules.

A. 11.0 for the Cisco 7000 platform and 21.0 for the Cisco 7500 platform.

• CIP Virtual Interface
• CIP Internal LAN
• CIP Internal Adapter
• External Communication Adapter (XCA) Major Node
• Switched Major Node
• Source-route Bridging the interface to CSNA

A. CSNA is dependent only on ring groups and bridge groups of physical interfaces. CSNA does not open service access points (SAPs) on physical interfaces.

A. The CIP Virtual Interface allows you to share internal adapters between CIP physical interfaces. It is always port 2 of slot x (for example, CIP slot x/2).

Note: The Channel Port Adapter (CPA) does not have a virtual interface.

A. The CIP Internal LAN is used to set up bridging to Cisco Systems Network Architecture (CSNA). It anchors the Internal Adapter configuration and has a maximum of thirty-two Internal LANs that may be defined per CIP.

A. The CIP Internal Adapter is used to configure MAC, Logical Link Control, type 2 (LLC2), and external communication adapter (XCA) information. It contains the Relative Adapter Number (RAN), the local MAC address, and the LLC2 connection parameters (T1, Ti, and so forth).

A. Cisco Systems Network Architecture (CSNA) implementation is similar to an IBM 3172. CSNA requires only one address for the LSA, unlike CLAW for TCP/IP, which requires two addresses. You may want to define multiple addresses to the Cisco router for spares.

A. Yes, CSNA supports duplicate MAC addresses. However, each must be on a different internal LAN and each must have a different Relative Adapter Number (RAN).

A. Capacity goals:

• Maximum number of concurrent channel-attached hosts: 64
• Maximum number of internal LANs: 32 (This is limited by the maximum number of internal adapters, which is 18.)
• Maximum number of internal adapters: 18
• Maximum number of opened service access points (SAPs): 128
• Maximum number of Logical Link Control, type 2 (LLC2) sessions: 6000

A. Use these IOCP statements to answer the next five questions:

CNTLUNIT CUNUMBR=aaaa,
PATH=(bb),UNIT=cccc,UNITADD=((dd,256)),
        IODEVICE ADDRESS=(eee,256),CUNUMBR=(ffff),UNIT=gggg

A. CUNUMBR in the CNTLUNIT macro is a unique one- to four-digit hexadecimal number in the range of 0000 through FFFE; it is arbitrarily assigned to each control unit. UNITADD in the CNTLUNIT macro is a two-digit hexadecimal number in the range of 00 through FF, which specifies the unit addresses of the I/O devices that are recognized by the control unit. ADDRESS in the IODEVICE macro is a one- to four-digit hexadecimal number in the range of 0000 through FFFF, which specifies the device number.

If this configuration is new, try to make aaaa equal to eee. For example, make CUNUMBR=0D00 and ADDRESS=(D00,256).

Also, make dd in UNITADD start at the last two digits of eee. For example, make UNITADD=(00,256) and ADDRESS=(D00,256).

A. No. You may define 256 channel addresses, but the CSNA will support only 64 concurrent channel-attached hosts per physical CIP port.

A. Use UNIT=3172.

A. Use UNIT=3088.

A. Use UNIT=3172.

A. Use UNIT=CTC.

A. CUADDR= in the XCA major node is the ADDRESS parameter in the IODEVICE macro. If the UNITADD for I/O device is 00 (CNTLUNIT macro), and its corresponding ADDRESS ID is D00 (IODEVICE macro); then you must code CUADDR=D00 in the XCA major nodes, if an only if you are using I/O device 00 as the “device” parameter on the router configuration (that is, the Cisco Systems Network Architecture [CSNA] path device). Use a CUNUMBR from the range indicated by the ADDRESS=(eee, 256) such as x'D08'.

A. xx is ignored for external communication adapter (XCA) major node and, therefore, is insignificant for CSNA.

A. It depends. Remember that the UNITADD in the CNTLUNIT macro is a two-digit hexadecimal number in the range of 00 through FF, which specifies the unit addresses of the I/O devices that are recognized by the control unit. This is the parameter that you code in the router for device in the CSNA path device definition. hh is dependant on what is specified in UNITADD=((xx,nn)), where xx is the starting address and nn is the number of addresses that are allocated to this definition. For example, UNITADD=((00,16)) says the addresses start with 00 and end with 0F.

A. Use this diagram to answer the next four questions:

ESCON Director Switch E1

CHPID PATH=((18)),TYPE=CNC,SWITCH=E1
CNTLUNIT CUNUMBR=320,PATH=(18),CUADD=A,UNITADD=((20,256)),LINK=E4,UNIT=SCTC 
IODEVICE ADDRESS=(320,256),CUNUMBR=(320),UNIT=SCTC

A. Yes. However, this example could be misleading because E1 is interpreted as the SWITCH number as well as the outgoing port to the host, from the point of view of the CIP.

A. It depends entirely upon what is configured in the UNITADD=(20,256) definition. In this example, the I/O devices start at 20 and not 00, and go up to 256 or x'ff'. You need to use CSNA E10A 28 or any free address in this range that you have chosen.

• Advanced Communications Function (ACF)/VTAM V3 R4.1 or later for Multiple Virtual Storage (MVS)/ESA.
• ACF/VTAM V3 R4 or later, for Virtual Machine (VM)/ESA, VM/SP, or VSE/ESA.

A. CSNA is supported in all major Mainframes. However, customers should be aware that Hitachi, IBM, and Amdahl have their own coded implementations for IOCP or HCD.

A. The channel is a point-to-point link. There is no defined interface to which to send the ping packet. Therefore, there is no way to wrap the packet back to the interface and respond to the ping.

A. 4096, if you want maximum performance. If you are using a gateway statement, you should also configure any network whose path will permit it to 4096, for maximum performance.

A. Interface processors have two mechanisms to send status information to the route processor. One is the love note. This is a MEMD buffer header with no data area. It is useful for sending 16 bits of information. The CIP uses this to indicate to the Route Processor (RP) that it should check the status of the “carrier,” because something has changed. The love letter is a real MEMD buffer. It contains up to 4096 bytes of data (RP/SP) or 1642 bytes of data (RSP). This data contains the statistics that are associated with each connection. It provides some of the output for the show extended channel commands and the MIB.

A. The coding unit (CU) gets into the resetting event condition whenever a system reset condition occurs. The ESCON specification states that, when this condition is present, you reject any command that is not an x'02' read or certain model-dependent commands. In that code, the CIP also accepts x'04', x'e4', x'84', and x'03'. If the first command that is sent down for this device is not one of those, then the CIP presents a unit check and sets the resetting event indication in the sense byte.

A. As of Cisco IOS® Software Release 10.2(4.1), the CIP will send RIP routing information to the mainframe host. If you are running the RouteD process on the mainframe, the mainframe will be able to make routing decisions based on this information.

Note: In order for the CIP to send the routing information to the mainframe, you must use the broadcast keyword at the end of the CLAW statement in the router configuration.

A. TCP/IP for Multiple Virtual Storage (MVS) and virtual machine (VM) has its own version of RouteD. RouteD is RIP-compatible. With OS/390 V2R6 and later, the OMPROUTE application could be used for RIP.

A. For each CIP card:

• Total memory.
• Free memory.
• CPU utilization.
• Time since last reset.
• Daughter board information.

Daughter board information:

• Daughter board type (such as ESCON or bus and tag).
• Daughter board status (such as up or down).
• If the daughter board has sensed light on the fiber (ESCON), or the Op Out Tag line (bus and tag).
• If the daughter board has established a path with at least one channel.
• The number of times that the ESCON processor has recovered from an internal error.
• The number of code violation errors.
• The number of link failures due to various conditions.

For each subchannel:

• Device address.
• Device connects.
• Halt subchannels.
• Selective resets.
• System resets.
• Device level errors.
• Write blocks dropped.
• Last sense data sent to the channel.
• The CLAW configuration.

For each CLAW link:

• Number of blocks read, written, and dropped.
• Number of bytes read and written.
• Number of times a buffer was requested and none were available.

The channel MIB is available via anonymous FTP from ftp.cisco.com:

SNMP	FTP Directory	Files
SNMP v1	/pub/mibs/v1/	CISCO-CHANNEL-MIB-V1SMI.my Files that start with “CISCO-CIP”.
SNMP v2	pub/mibs/v2	CISCO-CHANNEL-MIB.my Files that start with “CISCO-CIP”.

A. ESCON Multiple Image Facility (EMIF) is a mainframe I/O software function that allows you to share one ESCON channel among multiple logical partitions (LPARs) on the same mainframe. An ESCON director is a piece of physical hardware that allows you to share one ESCON channel among multiple device control units or one device control unit among multiple ESCON channels.

A. This is the current Offload memory calculation:

(64 KB × FTP sessions) + (4 KB × Telnet sessions)

At least 8 MB is required. For SNA, at least 8 MB is required. In both cases, the total can not exceed 64 MB.

A. Offload extends the application interface calls to the CIP, but the application resides on the mainframe. The TCP stack on the CIP terminates the TCP session, so all of the TCP/IP protocol processing takes place on the CIP. Application programming interface (API, socket) calls are transported from the mainframe to the CIP across the channel. IBM TCP/IP V3R2 is the last release to support Offload.

A. SNA sessions are not visible at the level of the CIP. What the CIP sees are logical link control (LLC) sessions, which map one-for-one into SNA physical units (PUs). The MIB support will provide a lot of information about the LLC connections; and so, by inference, the user can derive the state of the SNA devices. However, the CIP passes through the SNA session data transparently.

A. The SNA support on the CIP will make it look like a 3172. In other words, the CIP appears to the mainframe as a set of external LAN adapters. Support is provided for physical unit type 2 (PU2), PU2.1, PU4, and PU5 just as if they were LAN-attached to a 3172. Neither the 3172 nor the CIP present a PU image to the mainframe. They appear as a set of external communication adapter (XCA) Switched Major Nodes to virtual telecommunications access method (VTAM).

A. If you enable the TN3270 Server functions in the router, you can Offload TN3270 and TCP/IP processing from the host and saves host CPU cycles.

A. Yes.

A. In the past, most people would use a minimum of 16. The CIP minimum is 2. If you only have one TCP/IP stack per partition, two should be adequate. However, it actually makes sense to define more than two subchannels, if you want to use these:

• Second TCP/IP address space for:
  • Testing new versions or other products.
  • Security.
  • Performance.
• Something else, like Interlink DECnet.

A. On older Multiple Virtual Storage (MVS) Systems, the SENSEID information returned from an RS6K shows that it is a 3088 model 61, and the 3172 shows that it is a 3088 model (x'1F'). If you see the Host message, IOS445I Physical device is inconsistent with logical, then you are on an older version of MVS. You might need to install a PTF for APAR OY67361 or upgrade your MVS operating system to a newer, IBM-supported release. APAR OY67361 introduced changes, which included new UIMs for the 3172 and RS6K device type. Subsequent to release of the PTFs for this APAR, the recommendation—which comes directly from this APAR—states:

“A parallel or ESCON-attached 3172 should now be defined from the Hardware Configuration Definition (HCD) as UNIT=3172.”

Because your device appears to the operating system as a 3172, it is also recommended that you define your device as a 3172 on both the control unit and on the I/O device definitions.

A. There will always be a large number of retries with the CIP. This is normal. The retries counter increments when the CIP runs out of data for the host to read and goes into retry. You would expect the number of connects to be somewhat larger than the number of retries. If the number of connects is quite a bit larger than the number of retries, it indicates that the host is having trouble keeping up with the CIP card.

A. There are two limits. There is a limit of 256 devices per adapter. This limit restricts the number of CLAW connections to 128. In addition, there is a limit based on the amount of memory that each CLAW device requires.

This is the formula:

n₁ + (n₂ × number of ECA adapters) + (n₃ × number of PCA adapters) + (n₄ × number of CLAW connections) + (n₅ × number of unique paths) = amount of memory required

• n₁ = 555776
• n₂ = 705824
• n₃ = 714016
• n₄ = 10912
• n₅ = 1072

Note: The above are approximate numbers. They can change as software is updated. This gives roughly eleven CLAW device pairs on a dual ESCON Channel Adapter (ECA) with only 2 MB of DRAM. (In fact, you will probably get a little more because some of the control blocks will be allocated from 512 KB of SRAM, but this is a conservative calculation.)

A. Yes, CIP can autonomously switch to any other media type.

A. The sense data denotes this:

Byte	Bit	Meaning
Byte 0	0	Command Reject
	1	Intervention required
	2	Bus-out check
	3	Equipment check
	4	Data check
	5	Overrun
	6	Reserved
	7	Cancel
Byte 1	0	Resetting Event
	1	Reserved
	2	Reserved
	3	Reserved
	4	Reserved
	5	Reserved
	6	Reserved
	7	Reserved

CIP 4, hardware version 5.0, microcode version 21.10
Microcode loaded from flash slot0:cipp2110.bin
Controller Sync: 4 timeouts, 4 resyncs 0 failures, 0 max phase count

The Route Processor (RP) sends commands to the CIP via the Switch Processor (SP). The SP gives the CIP 12 µs to respond. If it does not respond in time, the RP counts a time-out: there were four of them. The RP then retries the request; part of the retry process is to make ensure that the CIP and the RP are synchronized and that the CIP is responding to the correct request. There were four resynchronization attempts. Most of the time, these occur when the RP is retrieving error counters, which it does once every 10 seconds.

The RP will retry each request 16 times. If the 16 resynchronizations fail, then a failure is counted. There were no failures counted in this example.

You may also see an 8010 error message, depending on the type of request. For some requests, the RP just ignores the 8010 error.

The max phase count reflects the maximum number of times that the resynchronization had to be retried. In this example, no resynchronizations were retried.