Chances are you want your mainframe on a network of some sort. Maybe not the Internet, but back in course one, we talked about all of the PCIe adapters that a system can have, and chances are, some of those are OSA or Open Systems Adaptors. These are the network cards for the mainframe and they come in a wide variety of flavors suited for copper and fiber links from gigabit to 10 gigabit, and even up to 25 gigabit speeds. On z/OS, TCPIP is handled by the TCPIP started task. So it starts up and stays up generally until the cyst Prague or another operator takes it down. Generally, we refer to this started TCPIP task as a stack, the TCPIP stack. Before we get into the TCPIP stack, I want to take just a minute to dive into the concepts of IP addresses and sockets. If you're already familiar with this from another platform, the next minute and a half probably won't be all that new to you. So you can go and check your email or scroll around a little bit if you don't want to hear it, I won't be offended. Typically, a system on a network has an IP address or two. You'll even see some z/OS systems that have five, six, even more IP addresses. An IP address is a 32 bit address represented in dotted decimal like this. So it looks like four numbers each between zero and 255 with dots in between them. That address is a lot like an address for a house. With that address, we can find the server on the network or even the Internet that we want to talk to. I'll also mention that the 32 bit address is what's called an IPv4 address. There's a constant concern that will run out of IPv4 addresses. So there's also IPv6. This is much longer, it's a 128 bits and it looks like this. It's longer, it uses Colin's data dots, and it's represented in hexadecimal. So there's a couple letters in there to. Many more possible combinations. You won't see an IPv6 address everywhere just yet, people are still trying to ride out IPv4 for as long as possible, but in case you see an IP address that looks different than what you are expecting, that might be it. Anyway, the IP address, be it IPv4 or v6 gets us to the server, but a server can run on multiple programs. How do we get to the exact program we want? Those programs operate on what's called a socket. So you've got the IP address, colon, and then the socket number. All that goes over the network. Once a socket connection is established, the program can start issuing reads and writes on that socket to get or put data. Then when it's done we close that socket and life goes on. So we have this TCPIP profile, and within it we find a number of statements that build our configuration step-by-step. For example, you'll probably find devise link and home statements. The device statement states which network devices we want to use, the subsequent links statements say how we want to use them, and then below that in the home statements, that's where we set the IP addresses and net masking and all that stuff. Now, this is a fairly simple setup right here. There's three devices, three links statements, and three IP addresses. It can get quite a bit more complicated where it's not always a one-to-one-to-one mapping. I just don't really think that we're making this complex for no reason. Now, let's talk about that VIPALNK1 entry. That sounds interesting, and yeah, actually it is. So typically an IP address should only be defined in one place because there needs to be no confusion about where that IP address is. Typically, if we tell someone that a server is at a specific IP address, and that server goes down or there's a network problem, people can no longer reach that server. That's when they start calling you and getting angry and nobody wants that. What a VIPA does, is it defines an IP address to the system virtually. Take a look at this diagram. We've got two LPARs each defined with the same VIPA in addition to their own unique IP addresses for each network link. This is a fairly standard high availability setup because if one LPAR is the primary, and also device one goes down for whatever reason, that VIPA is still reachable through the second pathway at the same IP address or even if the whole LPAR will go down, the backup takes over. It's got two links as well. So you can actually lose three links and an LPAR and the service is still available at that VIPA address. Pretty cool. For that reason, you'll often find critical services associated with a VIPA instead of an IP address that's hard coded to a single link or device. The virtualization and architecture of z allows for some pretty crafty ways of implementing highly available and highly performance solutions that you just won't find anywhere else, especially in distributed environments.
