Data Centers

The word data center is not too old for the IT industry. But, within a small

timeframe, many technologies have come up to solve the growing needs of data

centers to make them a better and economical place to keep your valuable IT

resources. Two trends that are driving the data center market are consolidatoin

and centralization. Most organizations are trying to centralize their IT

resources into the data center, because managing a centralized infrastructure is

easier than a distributed one. However, this in turn has put a lot of pressure

on the data center itself, making it much more complex and difficult to manage.

That's where consolidation plays a key role. We'll start with the key trends

addressing the consolidation challenge.

Virtualization



For most CIOs, virtualization doesn't need an introduction. In fact,

consolidation and virtualization go hand-in-hand, since the latter is the

enabler for the former. To make things more clear, consolidation means merging

your multiple resources into one. Let's say you need to deploy five applications

for your organization, and require two hardware servers for each (primary and

fail-safe). Consolidation says that instead of having ten servers keep two (for

failsafe) but more powerful servers and run all apps virtually on them. This not

only saves you the server hardware cost, but also cost of RealState, RackSpace,

electricity, and mgmt. The other good point about virtualization is that one can

recover any corrupted machine in no time, as the complete settings can be backed

up as a virtual hard disk or a virtual appliance. Products like XEN, VMWare, and

the upcoming MS Hypervisor are enabling such virtualization and making

deployment a piece of cake. With multilevel failsafe and great monitoring tools,

virtualization is becoming the key driver for the consolidation market.

Multicore/Processor



One prerequisite to implementing virtualization is that you need powerful

hardware servers. This essentially means servers with lots of processing power.

If you go back to our previous example of consolidating ten servers into two,

then, if all our services are eating up to 50—70% of the processing power, how

would you consolidate them with virtualization? How much powerful machine will

be required? Here comes the role of multicore processors. You can build a server

with let's say two quad core processors which means having a single machine with

eight processor cores. And that's not all. These multicore processors also

understand virtualization natively (Intel VT and AMD Pacifica), so you can

dedicate one core to one virtual instance and you will get near-native

performance for the apps.

IP KVMs



We have talked about the technologies driving the data center consolidation

market. Let's now talk about the technologies that had driven the Data center

Centralization market. The first one is indeed IP KVMs. They have made remote

access of servers possible and that to till the BIOS level. Not only this,

nowadays we have IP KVMs with which one can do a cold boot to the servers from a

remote location and all that's needed is either a web-browser or a tiny agent

that connects back to the KVM over a public IP or a VPN and gives near physical

access to the data center.

Which of the following have you already deployed in your data center?

Remote Infrastructure Management



When we talk about RIM, we are not just talking about accessing and managing

the servers from a remote and centralized location. Rather we talk about

accessing everything including temperature controls, power supply, security,

access, etc, centrally. To get such functionalities, different vendors are

providing devices such as intelligent racks and temperature controls. Such

devices are capable of providing mgmt interfaces over Web browsers which can be

accessed over public IP or VPN. Not only this, such devices can also send SNMP

alerts in case an error occurs on power or temp control. So if the temperature

of your data center increases to a threshold you can get an SMS alert and then

can solve the problem remotely using the remote mgmt interface of the cooling

system.

Where is it heading?



Many innovations took place in the data center market last decade. They are

still happening and that too at a greater pace. So now let's see where are we

heading with these new innovations in the next couple of years.

Self-healing data centers



Even after having so many innovative and automated technologies to monitor

and manage your data center remotely and centrally, there's a lot of human

intervention in the whole system. And wherever there is human intervention,

chances of errors increase considerably. In an environment where we need six to

nine 9s of reliability, it can cause problems in the long run. Nowadays,

technologists are looking forward to creating systems for data centers that

would require minimum human intervention. We will not say that self-healing data

center is something which is not at all possible today but the features and

adaption are very less. Today we have systems which can run some predefined

scripts and activities in case of a certain SNMP alert. So, let's say, the

processor utilization of a certain server has been raised to a threshold level

then there are systems which can automatically initiate a script and migrate

some of the processes running on the system to some other fail-safe system. This

much of self-healing is still possible but here we are talking about the future

where in case of physical failure, a failed disk array or a failed processor, a

set of robotic arms can get into action and replace the faulty devices with the

new ones and can automatically raise an alert to the hardware vendor. This might

sound a piece of fiction but believe me researches have started working on these

lines and we might see such products and services soon.

Workstation consolidation



This is something very interesting and has a great potential in the data

center market. Essentially workstation consolidation means consolidating all

your workstation into your data center and managing and provisioning them on a

need basis. Here the users only get a thin client or a KVM connection to their

desk and the actual machine is hosted in the data center and that too virtually

which they connect from anywhere and get resources as per their requirements.

When the requirement is fulfilled the resources are again released to the data

center which can then be allocated to some other user. This trend has already

started picking up in BPOs where they have to re-allocate resources on a daily

basis. Another form of workstation consolidation is app streaming, where,

exactly as the first case, the users only have access to a thin client or a KVM

interface. But instead of a whole system only the app which he/she wants to run

is streamed on his thin client. Citrix already has products on these lines and

MS is also coming up with such products in its upcoming version of Server OS.