A common phenomenon that has been happening across organizations is the
explosion in corporate data. Gartner Group has predicted a 650 percent growth in
IT data over the next 5 years, with businesses facing the pressing need to
deploy increasingly powerful and scalable enterprise servers to manage. Along
with this would come additional challenges like managing the power consumption
of these servers and the space they would occupy. Also the already existing
problem of server sprawl, prevalent in many corporate data centers where
numerous, under-utilized servers suck in a lot of space and resources, may
aggravate and call for consolidation. We look at some of the innovation
happening in the server space, that are aiming at creation of more efficient,
powerful and energy efficient servers for enterprise data centers.
Buy memory instead of server
A typical x86 platform server, with its 30 year old architecture, has memory
capacity tightly locked, alongside the server's processor. With the shooting up
of the memory-intensive workload, the only option traditionally resorted to, by
the IT managers is to get in more servers to resolve the problem, which inturn
directly pushes up the power and management costs of the data centers. How about
buying memory that you need without buying new server and the related hardware?
IBM has tweaked the traditional x86 platform and has come out with its new
x86 system — eX5, wherein they have decoupled the memory from the chip, using an
independent memory scaling technology, called MAX 5, thus enabling IT managers
to scale memory in their data centers without having to buy new servers. eX5
servers have incorporated IBM's fifth-generation Enterprise X Architecture
chipset with Intel Xeon processors, which is enabling them to double the amount
of memory available in an IBM server when compared to other Intel-chipset-based
servers.
The X86 blade servers generally come with 12—16 DIMMs and if these are used,
then IT Managers have to buy a new server. With IBM's eX5 blade systems one can
add an additional 24, while rack systems will enable adding up as many as 32
DIMMs. With this upward scalability of memory for racks and blades,
organizations can now support more virtual machines, leading to faster database
performance and greater server utilization. This technology is said to allow
running 82% more "virtual servers" for the same license costs and also
drastically bring down the middleware and application costs. The eX5 portfolio
includes three systems: a four-processor IBM System x3850 X5 server, a
two-processor System X3690 X5 server and the BladeCenter HX5 blade server.
Servers for data-intensive services
There are a lot of applications and services that require processing of
large amount of data and real-time analysis of large amount of data, for
example, to control smart grids or analytics for financial markets. IBM is
coming out with new Power 7 enterprise servers, based on its latest Power7
processor in March, in India. The Power7 processor has upto eight cores, each
core being able to run four threads and with the ability to run 32 tasks
simultaneously. Chips being made using the 45-nm process technology have 4
times the capacity of previous generation power systems, four times the
virtualization capability and 2x-3x times the energy efficiency. Some
interesting technologies included in the processor are:
Intelligent Energy — An energy efficiency technology that enables
increasing or decreasing of the processor clock speeds on the basis of system
usage or thermal conditions, either on a single server or across pools of
servers, the advantage being, when the workload is demanding, it can be
provisioned for by quickly speeding up the processor cores and if the workload
is lesser, turning off of some cores or slowing the clock speed, thus saving
energy.
Active Memory Expansion — A feature that makes the physical memory on
the system appear to the application as if it were up to twice as large as it
actually is. This enables a partition to have access to more memory, thus
facilitating the usage of software such as SAP applications, that require large
amount of memory or for virtualised environments where greater amount of memory
is a necessity. TurboCore mode can be utilized for database or other
transaction-oriented workloads. Switching to this mode will lead to utilization
of up to 4 cores per chip with twice the cache per core and at a greater speed,
just like having 4 sprinters. This saves costs of the clients, like software
costs for those applications that are licensed per core, such as databases and
some application servers.
Virtualize servers: save space and power
HP's Integrity servers based on Intel's quadcore Itanium processor 9300 series
previously called Tukwila bring significant increase in performance along with a
range of new reliability, scalability and virtualization features. It runs eight
threads per processor, through enhanced Intel Hyper-Threading Technology. It
runs eight threads per processor, more cache, up to 800% the interconnect
bandwidth, up to 500% the memory bandwidth, and up to 700% the memory capacity,
using DR3 memory.
The Itanium 9300 processor employs the second generation Intel Virtualization
Technology to improve performance and with its Intel 7500 chipset it can
directly assign I/O devices to virtual machines. Added to this is the HP Insight
Dynamics — Virtual Server Environment, lifecycle management software for HP
Integrity servers. It allows IT managers to quickly enable, plan, configure, and
automate physical and virtual resources. With virtualization being facilitated
this way, direct savings in energy, space and time required for maintenance can
be observed.
HP's Converged Infrastructure
This is a solution to check the IT resource sprawl. It involves creating
management tools, policies and processes, that enable separation of applications
from their underlying physical assets and unifying these apps into the virtual
resource pools. It includes end-to-end virtualization, that virtualizes all
resources in the data center: compute, storage, networking and power and
cooling. This bestows the freedom to move application environments anywhere,
anytime, thus enabling optimization of every data center resource and quicker
response to business requests. This converged infrastructure is based on
technologies that corporate data centers are already using, virtualization and
blades.
Servers for efficient power monitoring
Fujitsu's PRIMERGY range of servers are delivered with Server View
Management Suite, which enables easy setting up, management and recovery of the
IT infrastructure. Also the iRMC2 chip (integrated Remote Management
Controller) integrated on the motherboard adds remote management functionality
with the basic system management functions.
The iRMC2 chip and ServerView together perform certain power management
functions, like remote, power management controls, enabling power on and off
and shutdown for physical servers as well as on virtual servers, and power
monitoring that enables analysis of the power consumption of a single or
multiple servers. The iRMC2 chip enables controllingthe power consumption and
the scheduler function allows switching between two modes.
The first is Minimal Power Consumption where CPU power is constrained to the
lowest minimum used by the CPU type, and the second is Best Performance mode
where the entire CPU performance is made available for the OS.