Class:       Advanced S-Series TCP/IP Overview
Sponsor:     HP Westshore Tampa Office
Instructor:  Jim Hammond
Date:        2nd October, 2002

Thanks to John T. Rodda for these terrific notes!


This day-long class was designed as an overview of S-Series telecom subsystems with an emphasis on web-enablement via TCP/IP. The material was culled from several other HP Nonstop Systems Classes minus the lab practicums. The class was designed to expose potential students to material covered in related HP Nonstop course offerings and generate interest among local NSK users in this training. A very elaborate and extensive book was given to all attendees, and there is definitely no need to copy and/or paraphrase any of its contents here. I have instead included comments made by the instructor in support of the text.

Intro to S-Series Communications Subsystems

  1. Almost everything concerning datacom on the S-series is performed via TCP/IP.
  2. Ethernet failover has been added to OS G06.10. This provides fault tolerance at the adapter level.
  3. The K and S series systems are of very different architecture, except when it comes to porting applications.
  4. The architecture has been designed to interface with SONET networks and ATM (Asynchronous Transfer Mode) switches. ATM switches assign addresses dynamically.
  5. Permanent virtual circuits can be established with fixed addresses.
  6. The S-Series has an excellent LAN router. It's interesting to note, however, that none of the class participants use it for this.
  7. TCPSAM (Transmission Control Protocol Socket Access Method) - NSK applications link through this. Please note that it is not part of the data path, and is used only for open and close operations.


TCP/IP Protocol Suite

  1. UDP (User Datagram Protocol) - Used for SWAN (Servernet Wide Area Network) connectivity. Great for downloads.
  2. ICMP (Internet Control Message Protocols) - Command message transmitters. A good example is "PING".
  3. Note that ATM has long (20 octet) addresses.
  4. All TCP/IP processes are setup via SCF.
  5. Logical Interfaces (LIF) within the SLSA identifies the physical access ports. Conventional and parallel libraries cannot be shared through the same LIF.
  6. A listener processes handles server requests not coming through LIF's.


Servernet LAN Access Systems

  1. SLSA (Servernet Lan Systems Access) is used to interface the S-series with LAN's. Note that it is pronounced "Salsa".
  2. Uses 10BaseTx - Twisted pair standard with RJ-45 connection. Please note that fiber optic lines are much more secure than twisted pair. They're harder to hack.
  3. SLSA/Ethernet is a contention-based protocol.
  4. Although Ethernet is not as efficient as a Token Ring, it is significantly less expensive.
  5. Generically-named processes: The S-series remembers the configs and invokes them automatically.
  6. The PIF is the actual physical interface port and is logically represented by the PIF.
  7. TCPMONs can all have the same IP address.
  8. An S-Series access list determines which processors can communicate with each other. Please note that this particular access list pertains more to productivity than security.


SLSA LAN Config for TCP/IP

  1. Note that no one on the S-Series should be running anything below G06 at this point.
  2. First start the TCP/IP process from TACL, then use SCF to add Subnet.
  3. Implicit routes to direct local traffic are performed automatically after starting Route.
  4. When reconfiguring a system to add or delete processors, you must first stop the SAC (Servernet Addressable Controller). If only altering, then this is not necessary.
  5. The IRDP attribute is off by default. Be careful when turning on, as you could inadvertently open the door to hackers.
  6. SNAX used to interface with the token ring.


TCP/IP PL Overview

  1. Up to 16 processors can be assigned to a particular IP address.
  2. Can access TCPMAN (management) through TACL, but it's more efficient to set this up as a generic process.
  3. Note that with the S-Series, doubling the number of processors will double throughput.
  4. Typically all TCP/IP configuration is performed through SCF.
  5. After a logical interface (LIF) failure, the backup LIF automatically becomes primary, and the primary becomes the backup. Be aware, however, of possible contention issues when the config is restored to its original form.
  6. Remember that only one listening process per processor is allowed. This could also result in port collisions in backup situations.


WAN/SWAN Overview and Expand IP

  1. Existing NSK standards and processes are used to accommodate the SWAN environment.
  2. The WAN Wizard Pro helps administrators to easily configure the environment.
  3. Note than in the SWAN environment, the MAC addresses of the physical locations remain the same when the associated modules are replaced.
  4. The NCP (Network Control) automatically calculates the best path in multi-node systems.
  5. Note that Expand is the only network architecture that is part of the OS.


Miscellaneous


For information on NSK consulting and alliance partners, go to http://nonstop.compaq.com/view.asp?PAGE=TBU_Partners or contact Joy-Ellen Lipsky at joy-ellen.lipsky@hp.com.

For training and certifications, contact Francine Barr at Francine.barr@hp.com.

Finally, the presenter can be contacted at jim.hammond2@hp.com.



last modified Wednesday, December 18, 2002