|
Smaller, stronger servers
Powerful, efficient and compact servers with some sort of
intelligence and management capabilities in-built to ease data centre management
are the order of the day. By Faiz Askari
The
ongoing economic boom across verticals has sent IT heads into a tizzy over growth
inside their data centres particularly on the server side. Servers are considered
to be among the building blocks of a data centre. Enterprises are looking to
save money and enhance production by consolidating workloads onto more power
yet energy-efficient servers that can bring down the cost of operating a data
centre.
Virtualisation has engendered a good level of confidence amongst CIOs and data
centre architects. Virtualisation technologies separate software from hardware
letting users eliminate multiple physical servers and get the processing power
that they need from fewer boxes.
This brings about the need for virtualised servers interconnecting with each
other. Such technologies speed performance and ensure security. HP, IBM and
Sun have all looked at or developed these server-interconnection technologies.
Other technologies such as InfiniBand, the high-speed I/O switching fabric and
10G Ethernet, also are emerging or have emerged.
Suresh Menon, business manager, Business Critical Systems,
HP India Sales says, Today every enterprise needs a robust and agile IT
infrastructure. Real-time access to data and communications is the lifeline
of all businesses and basic IT infrastructure has now evolved into a mandatory
business requirement.
|
"Low
end servers will set the tone in the Indian market. These systems are
making a mark as they assume important tasks such as virtualisation and
are also having large amounts of memory as well as internal storage"
- Arnab Roy
General Manager
Marketing
Sun Microsystems India
|
Giving an insight about the technologies that are driving
this market, Arnab Roy, general manager, Marketing, Sun Microsystems India says,
The technologies that are driving the server markets currently are most
definitely the Unix servers.
However, the technological acceptance of low end servers is
greater than that of high end ones and hence their growth is much faster. Blade
servers have gained ground.
Roy adds, Low end servers will set the tone in the
Indian market. These systems are making a mark as they assume important tasks
such as virtualisation and are also having large amounts of memory as well as
internal storage.
|
"Performance
per watt is an important metric. As the number of data centres and density
of data centres increases, IT managers have to think about the cost of
operation."
- Surendra Arora
Head of the South Asia Enterprise Solution
Sales Group and World Ahead Program of Intel
|
Surendra Arora, head of the South Asia Enterprise Solution
Sales Group and the World Ahead Program of Intel highlights some technologies
that are driving the market of servers for data centres. He says, Technologies
such as multicore, virtualisation and service oriented architecture are actually
driving this sector.
In the data centre
Data centres are evolving. Most companies that earlier had unstructured data
are ramping up facilities and there is a growing level of awareness among CIOs
and data centre managers as to how they can manage their infrastructure optimally.
A lot of R&D is going on with regard to optimising server power consumption.
This is happening from the processor level all the way up to disk and cooling
technologies. Processors will have the ability to slow down or reduce power
consumption when not being taxed by heavy loads. Similarly cooling fans which
operate at differential speeds depending on a servers requirements reduce
power usage. Viswanath Ramaswamy, country manager, eServers xSeries & Intellistations,
IBM India/South Asia says, Earlier we had to sell the concept of blades
to the client. Today instead, it is the client calling us to enquire about blades
or solutions to reduce the energy requirements of his data centre. So I think
ease of management, consolidation and energy efficiency are also factors driving
buying decisions and not price alone. The concept of TCO has finally begun to
take root and is no longer alien to most data centre customers.
Arora says, Server consolidation is a key growth driver. As such CIOs
are looking for dual or multiple processing platforms to optimise the data centre.
Blades are also driving the market on account of the scalability that they offer.
Ramaswamy says that IBM is taking this a step further with the launch of Power
Executive software that it is shipping free with almost all of the companys
rack and blade servers. This software is designed to work with Intels
Demand Based Switching (DBS) and AMDs PowerNow! to reduce the amount of
electricity consumed by servers when in an idle state. It also goes a step further
by allowing users to cap power available to a server. Thus these technologies
are helping remove the cooling or power availability bottleneck that many data
centres face, allowing for the purchase of additional servers.
Data centres are growing in numbers as well as complexity, providing a huge
business opportunity for both solution providers as well as vendors. Menon adds,
The growth of the data centre market in India is attributed to the move
towards centralisation and consolidation at the business process, application
and infrastructure level.
Performance per watt
Performance per watt was a new thing until last year. Indian customers today
are looking for data centres that are much more optimised for space, power consumption
and performance. Customers today have adopted a three dimensional view of evaluating
not just the cost of acquiring infrastructure but also the cost of running it.
With the emergence of dual core architecture, customers are getting more bang
for their buck. Ranjan Chopra, CMD, Team Computers says, At the same time,
investment in power and cooling are less, resulting in lower cost of ownership
and running expenses. Arora adds, Performance per watt is an important
metric. As the number of data centres and density of data centres increases,
IT managers have to think about the cost of operation. Intels Xeon 5100
(Woodcrest, dual core) and 5300 (Clovertown, quad core) series are widely regarded
as the most energy efficient processors in the world. Intel is the only
vendor to offer quad core servers today. The Xeon 5300 series maximises energy
efficiency and gives the most performance head-room.
According to Ramaswamy though Performance per watt is an important barometer
abroad, the concept is still in its infancy in India. However he also said,
We see this as a major driving force in future purchases, especially as
the trend of offshoring data centres to India continues. The ability to get
maximum performance out of limited or minimal power is already a major decision
criterion for some of our EDA clients in India and we expect this to percolate
to other industries in the coming months.
Putting this across in other words, customers are now looking beyond price and
performance to performance per watt per rack unit of space. Roy adds, Sun
has place SWaP (Space, Wattage and Performance) which helps CIOs to calculate
and make comparative notes of the performance of the server that they are going
to purchase. When we say that our servers are the most energy efficient, we
have gone ahead and asked customers to try it out themselves.
Agreeing that this is an emerging concept in India, Menon says, Enterprise
customers are evaluating this trend very closely. Performance is still the key
parameter, but Performance/Watt is becoming an important measure as it directly
starting to show as operational expense. Having the lowest Wattage per
server does not matter. It is the combination of high performance and low power
consumption that is needed.
Roy says, Efficiency in every aspect of an organisation is necessary,
and Sun Fire servers with CoolThreads technology consume one-third less energy,
generate less heat, and take up one-half the space in the data centre while
delivering twice the performance. Sun avoided approximately $24 million
in energy and systems costs last year by using its own technology.
Arora adds, CIOs need to start thinking about refreshing their data centres.
Last year, Intel IT proactively retired legacy single core systems. The team
saved nearly $9 million, increased performance by 80 percent and reduced power
consumption by 40 percent.
A Gartner study shows how CIOs need to quantify and
manage the problem of infrastructure growth. Many methods can be used in
large data centres and all of these should be employed in a coordinated
manner.
- Improve the use of existing equipment
because this will delay the need to move to high-density systems.
- Balance the technical requirements of
new equipment with power and cooling specifications
- Configure data centres to balance air
flow between hot and cool zones and create special high-density zones
for more demanding infrastructure. Install chassis that use chilled
water and explore other solutions as they become available during the
next couple of years.
|
Management and maintenance
Roy says, Sun Managed Services provides a perfect solution for managing
and maintaining data centres. Sun Managed Services is a broad portfolio of dynamic,
heterogeneous IT infrastructure and management capabilities aimed at helping
customers enhance the business value of their IT investment through improved
operational efficiency and service levels.
Blades and virtualisation when combined help reduce age old problems experienced
by even the most proficient of IT professionals. Chopra says, Problems
such as deployment of servers, patch management and getting by with less physical
hardware can be overcome with the help of virtualisation.
Menon says, HP has simplified the management tools. Consolidation and
virtualisation which provide great cost reduction and flexibility can add to
the management requirement. Also most data centres have multiple operating environments,
such as Unix, Windows, and Linux. Management tools from HP are browser
based, requiring little or no training and are common across all categories
of HP servers and storage and are available out of the box to manage HP UX,
Windows or Linux environments. These can be easily integrated with enterprise
management tools such as the HP OpenView suite of products.
- Maximise compute density, so you can pack
more performance into a smaller space.
- Optimal energy efficiency.
- Servers are also optimised for reliability
and serviceability, features that are essential in minimising the risk
of consolidating dozens of legacy servers and applications onto a single
server.
|
Server consolidation
The complexity of the IT environment and systems has seen consolidation in the
enterprise data centre. Data centre operations have been consolidated across
industries due to falling costs.
Menon adds, Consolidation is gaining momentum not only in large enterprises
but also the SMB segment, though at different levels. The need for additional
capacity is now being contested with consolidation and centralisation leading
to better resource management and lower downtime. In the present environment
where businesses are trying to bring down costs, consolidation makes sense on
a broad level.
Agreeing to the fact that server consolidation is an emerging area, Roy says,
There are apparent hidden costs to any infrastructure, which is becoming
apparent for many enterprises. For e.g. with respect to x86 servers, there are
issues such as poor server utilisation rates at 20 percent and server sprawl
that is costly to manage, maintain, power and cool.
Arora adds, Overall TCO decreases as servers are consolidated. The exact
benefits depend on the application. An Intel IT study has found that for every
100 servers consolidated using the quad-core Xeon 5300 series, IT departments
can save over $75,000.
According to Roy, Physical consolidation is one step in increasing resource
utilisation and cutting infrastructure costs. However, any IT manager also wants
assurances that this consolidation will minimise disruption in the data centre.
Sun x64 systems address these issues head-on by enabling x86 server consolidation
through virtualisation.
Chopra highlights that consolidation helps reduce power, cooling and overall
infrastructure costs. By consolidating servers, less time is required
for managing servers and data centres. This in turn allows organisations to
invest more time and focus on critical areas that will directly impact business
productivity, says Chopra.
- Optimises the need for space.
- Cost of implementing virtualisation is
a fifth of the investment involved in purchasing the actual product.
- Pooling of data as per specific profiles,
similar to application storage with minimal sharing or data duplication
or overlapping.
- Disaster recovery is easier, in case there
is a total systems failure.
- Diverse operating systems and applications
can be consolidated quickly and easily.
- Enhanced availability and security is
also possible in a more comprehensive manner. Software faults and digital
attacks are isolated within each virtual partition, and fail-over partitions
can provide an easy, cost-effective approach to tailoring availability.
- Server virtualisation enables legacy applications
and their existing OS versions to be migrated without modification onto
virtual partitions--extending the useful life of legacy applications
at relatively little cost and with less risk.
|
V is for virtualisation
Virtualisation is one of the most dynamic trends in the market. Companies are
concerned about maximising their investment as well as increasing server utilisation.
Menon says, Virtualisation is about pooling resources. However, we believe
that customers must consider more than just pooling of resources if it has to
be useful in a business application environment. Thus we believe that a Virtual
Server Environment is all about virtualisation integrated with high availability,
pay-per-use models and superior management capabilities.
Virtualisation is on the rise, even in the x86 server market. Traditionally,
customers are used to running a single application on an individual server.
With utilisation levels languishing at less than 20 percent for x86 hardware,
the boom in processing power with the advent of multi core chips can only exacerbate
this wastage. Hence the growth of virtualisation with IT managers choosing to
run VMware, Microsoft or Zen OS to deploy multiple virtual machines each with
its own application on a single physical server.
With processing power and memory addressability growing by leaps and bounds,
the only remaining bottleneck to virtualisation is I/O. Ramaswamy says, Even
if your server can support multiple virtual machines in terms of processing
power, having three or four 1GB Ethernet connections for eight to ten virtual
machines can be a limitation. The adoption of technologies like InfiniBand
or 10G Ethernet can be used to address this problem. IBM has already announced
products on the above standards in its blade and rack server lines and one of
the major differentiators of the companys Blade Chassis is the high I/O
bandwidth availabilitya must if you are looking for a successful virtualisation
deployment.
There must also be tools and technologies which enable automation and integration
with business applications like those from SAP and Oracle. It is only when all
of these are addressed holistically that virtualisation will benefit a business
organisation.
Highlighting some problems that a customer usually comes across in India,
Roy informs A sudden surge in a companys customer base and transaction
volumes is one. One of our banking customers saw a surge in their transaction
volume from 50,000 transactions per day to 2.5 million transactions per day.
Another big issue ahead of customers is to scale up as per the perceived market
demand, without investing in additional IT infrastructure.
Moreover, the development of online infrastructure, for robust and speedy
transactions, scaling up the operation and reaching out to a bigger audience
is a situation faced periodically that is addressed to IT engineers, and they,
in turn, provide solutions such as ERP implementation as well as pooling. This
is an aspect of virtualisation, which helps, in multiple applications as well
as in reducing the footprint of physical systems and putting dedicated resources
onto a task.
Currently, server virtualisation is most common on proprietary hardware in data
centres, but over time, it is likely to become equally popular among smaller
businesses. It will allow them to simplify deployment and improve reliability,
availability, and security. Arora says, Virtualisation lets you turn a
physical server into multiple virtual systems, or partitions, allowing
multiple operating systems and applications to run independently on a single
platform. Intels virtualisation technology provides hardware enhancements
built into the companys next generation server platforms, improving the
stability of what today is a software-only solution.
Applications hosted at a data centre, big or small, must be accessible at all
times. This translates into a requirement for robust server architecture. Features
such as component level redundancy and in some cases, server level redundancy
help make sure that the client always has access to data, ensuring business
continuity.
Robustness is a confluence of RAS-Reliability, Availability and Serviceability.
This is an integral aspect of any server architecture. Vendors have some offerings
that sport mainframe class RAS features, which have been brought down to the
Unix environment at affordable prices. Roy says, We have a rich history
in terms of features like dynamic system domain, ability to run mixed generation
clock speeds, ability to service a running system without a reboot of the OS
and extensive virtualisation.
Arora adds, Robustness of servers is extremely important for any data
centre environment. That is why Intel has taken a platform approach to server
designs. For example, Intel I/O acceleration technology improves data throughput.
We also offer demand based switching which reduces average system power consumption
and potentially improves system acoustics.
Calling server robustness the most critical requirement in running mission critical
business applications in a data centre, Menon adds, The server and solution
architecture must be designed to handle any eventuality as downtime can result
in tangible and intangible losses. Thus most customers choose servers and operating
environments with proven capabilities and robustness for running their mission
critical applications.
|