The mainframe is
an ever evolving platform. New features are constantly
being developed, existing components
are getting upgrades, and its capabilities
are always being extended to meet
the world's business demands. As it evolves,
the mainframe gains some unique capabilities that you won't find anywhere else. I'd like to
cover a
few of those here. We'll start off with the
Intelligent Resource Director or IRD. The IRD function
optimizes processor and channel resource
utilization across LPARs within a single Z system. It groups LPARs into
an LPAR cluster, allowing WLM to manage
resources across the entire cluster not
just in one single image. IRD has the following
characteristics. First, it allows for more
flexible LPAR CPU management. Next, thanks to Dynamic
Channel Path Management, DCM, phicon channel bandwidth which is primarily used
for connections to storage devices can
be moved between disk control units to meet
demands as they change. Lastly, Channel Subsystem
Priority Queuing allows I/O requests in
the channel subsystem to have priorities
assigned to them. This allows for
even greater control and flexibility when we have important business
objectives to meet. On the network side of things, hyper sockets provide
high-speed connectivity between servers in
the same physical system. These connections are entirely virtual so there's no need to run
cables or hook connected servers into a networking infrastructure. You define a hyper
socket and then virtually connect it to z/OS, Linux or z/VM systems. You can even mix and
match. To the OS, it looks like
a regular networking device. They can now communicate as
if they were plugged into a super high-speed network switch except it's all virtual. Up next,
hyper dispatch. So when we virtualize the LPARs, when there's work to be done, the system
looks at the pool of available processors and lines of that specific unit of work within available
processor
that can handle it. This is all fine and good, but in some workloads, we gain performance by
ensuring that the work is dispatched to the same processor as frequently as possible. So
there's no weight while we figure out which
processor the work should go on and we also
benefit from reusing relevant data left in the cache
from previous operations. It's like landing at
the airport and renting a car. If you have to go
find the right car each time and adjust the seat and the mirrors and figure
out how to pair the Bluetooth to your phone
in it, that takes time. Hyper dispatch is like having your favorite car park right outside waiting
for
you all the time. Hyper dispatch really
helps WLM and keeping consistent response times
and definitely helps once you start scaling up systems with lots of processors, with lots of
cores. Now, setting up
mainframe hardware can be a fairly complex task. There are just so
many definitions and bits of
information you've got to manage correctly in order
to add storage, for example. Fortunately, zDAC,
the z Discovery And Autoconfiguration aims to help make that a little bit easier. It can detect
certain devices on phicon and Storage
Area Networks and streamline the process
of configuring the HCD or Hardware
Configuration Definition. This helps to simplify I/O configuration and reduce
complexity and setup time. Virtual flash memory. Now, earlier, we talked about storage and
how virtual
memory means that the memory for a program might go from on-chip memory out to DesD.
Well, there's actually
an in-between step and that's where
virtual flash memory comes in. This is storage class
memory that lives in the same physical box as
the memory and processors but can be configured to provide additional support for
programs that take a noticeable performance
hit when they page data like DB2, Java, and the coupling facility. Then, there's
zAWARE. As you know, the mainframe is
constantly gathering data. Not just data for
business transactions, but information about what's
going on behind the scenes. Network configuration changes, LPARs coming up and going
down, software levels changing,
new user ID is getting created, and nobody has time to go through all of those log files
line by line by line and inspect every single thing that
happens during the day. Fortunately, zAWARE
which stands for z Advanced Workload
Analysis Reporter does. It watches everything going on and compares what it sees with prior
system data highlighting anomalies and suggesting
courses of action. It does all of this
in near real-time. So you can even use
zAWARE's interface to diagnose the cause of past or
current anomalies. As you can see, there's a lot of uniquely z features
that you can explore. There's always more in the works. Now, if I told you that you
needed three LPARs set up in a cluster with virtual
network devices between them, you know exactly what
features to look into, right? Well, I sure hope so
because it's pop quiz time. How did you do? I hope you guessed IRD and hyper sockets
because
that's the right answer. We would use IRD to set up
a cluster of LPARs and then hyper sockets provides
the virtual network and virtual network interfaces
to connect those LPARs.