Supporting Hardware - Expert Consolidation in Oracle Database 12c - page 72

a high degree of professionalism is expensive. Especially in large organizations where all major platforms are in use,
a great deal of time has to be spent to certify a given platform with storage, the operating system, and software stack.
Naturally some of this work can be reduced by limiting the number of supported platforms. The number of supported
platforms depends on the strategic direction of the company and also on the budget available. Certain vendors,
however, may not be able to comply with the platform strategy, but that is not a problem either. Support arrangements
can be put in place between the vendor and the application team to cover the nonstandard platform as an exception.
The vendor should then be persuaded to adjust its product to be in line with the corporate strategy.
One often-heard argument against limiting the number of supported platforms is that of the vendor lock-in.
Being supplied by only one or very few parties for mission critical systems can be either a big advantage (as the vendor
marketing will tell you) or a disadvantage (what the other vendors will tell you). There is no clear-cut answer to the
question of vendor diversity. This again is often down to managerial decisions and relationships with vendors, but the
Linux-operating environment at least gives you a lot of choice.
However, despite the strong favor of Linux in this topic, you should not feel motivated to rush and replace your
existing estate with x86-64 servers. During many site visits it has become apparent that the staff responsible for “their”
operating system (and the servers running it) often have strong feelings towards that platform. In addition to the
necessary skill that the administrators must possess to manage a platform, it is important to use adequate change
management when introducing something new. The concerns and worries of staff should be taken into consideration,
and the budget forecast should include training or new hiring as well.
After so much theory it is time to get back to the real matter: hardware! This section will start off with a question
of whether or not blades make sense and whether rack-mounted systems fit into your organization better. It will then
discuss the changes in the hardware space that happened in the past two years before exploring more advanced
aspects of the x86-64 platform.
Blades or Rack-Mounted?
Blades have established themselves as suitable alternatives to the classical 19-inch rack-mounted server for many
workloads. Blades are usually smaller than rack-mounted servers, but they also have fewer components and fewer
possibilities for expansion. Some of the components you would normally find inside a rack-mounted server will
be in the so-called blade enclosure. The small form factor for blades will allow a very high density, and when they
were first introduced, blades seemed ideal to reduce the space required in the data center. However, with the high
density comes a heightened requirement to cool the blades and enclosure to prevent them from overheating. With
the introduction of chips capable of adjusting their clock rate depending on workload, cooling becomes even more
important. Often the chips need to be in their comfort zone when it comes to their Thermal Design Power (TDP).
As soon as the processor temperature rises too high, it will clock down and run at reduced speed to prevent damage to
its components. Sufficient cooling of modern CPUs is therefore essential to stable CPU performance! Luckily the ever
shrinking manufacturing process for processors allows for a reduction of cooling requirements. Processor vendors
have realized that a power-efficient processor is a real necessity in the drive to cut cost in data centers.
There is no general industry standard blade enclosure, but you can expect most blade enclosures to
contain power supply units, networking equipment to support storage and client connectivity, and other shared
infrastructure. The introduction of Data Center Ethernet or Converged Ethernet allows vendors to use high speed
Ethernet switches for network and storage traffic, potentially reducing the number of cables in the data center. The
matured graphical management interfaces allow the administrator to perform lights-out management of his blades.
Graphical representations of the blade enclosure tell the administrator to see which slots are free and which ones
are occupied. Often warnings and alerts can be sent via SNMP (Simple Network Management Protocol) traps to
monitoring solutions.
A blade enclosure would be of little use if it were not for the blades. The little, yet powerful computers can either
be horizontally or vertically added into the enclosure. Depending on the model, they can either occupy a full slot or
half a slot. The terms full-width and half-width (or height) have been coined for these. Blades are highly versatile and
configurable. If your data center can accommodate them, they are definitely worth evaluating.