|
Data Center Builders
Introduction
Building new Data Centers is a costly, complicated and with overwhelming details
To build a new data center, the cost and the time it takes are serious issues, plus the
top executives may not see the cost as justified. We believe that using good management,
virtualization, intelligent building tools and automation would reduce the cost of building and the
running data center to a fraction of what is done today. Building data centers would requires
building the following:
• Management teams
• Planning
• Budge
• The physical structure - building, power, internet connections, ... etc.
• Networks (OS, servers, routers, ..etc)
• Network skilled staff
• Testing team (on/offshore)
Therefore, building data centers must be broken down to smaller and manageable and affordable
steps with contingency plan. The first important ingredient is team work where all parties
involved must have their views and participation in the Work Breakdown Structure (WBS) and
first major milestone. The new data center (facilities and sites) would involve literally
every IT departments and testing the new site is critical to the data center success.
Use Case:
Using Intelligence and Virtualization to Automate Building Data Centers
Description:
Is our claim a fancy or a con?
According to the internet, most companies have between 50,000 to 100,000 servers. The
actual cost of building a new data center with latest and the greatest will a run into
the $Billion and will take a long time to build. Most executives and stakeholders may
look at our claim which we would cut the bulding cost to fraction as a cone or
a fancy. Our attempt in this page is to present our approach and architect and the
world is free to use it or not.
Briefly Describe This Use Case
We believe that using good management, virtualization, intelligent building tools and
automation would reduce the cost of building and the running data center to a fraction
of what is done today.
Primary Actor:
Executives, stakeholders, Infrastructure directors and architects.
Goals:
To present a solid and sound approaches with an architect which removes any doubt about we are presenting.
Analysis
The first strategy would be “Divide and Conquer”. There is must be a pilot project
as a test ground and learning curve. The best start is searching and a planning of
the first manageable and affordable build. Once the plan is approved, then building
the pilot project should be completed and tested. This pilot project is the basis
for building the rest of the data center. Automation and intelligent tools such as
our Building Editor must be used to cut cost and speed the building processes.
Architect
This pilot project is the basis for building the rest of the data center. Automation
and intelligent tools such as our Building Editor must be used to cut cost and speed
the building processes.
Virtual Pilot Architect Diagram
The virtual network structure and tools factories must be intelligent and automated.
Plans and maintenance-troubleshooting , testing, reports, statistics and
DevOps would be used to create, copy-paste and clone virtual networks. Management
and documentation must run the networks. Reports and statistics would be used to
evaluate the running networks.
Virtualization, Virtualization, Virtualization and more Virtualization:
We are talking about:
• Virtual Servers (we define them as
Sub-Servers Objects in OOD Model
)
• Virtual Containers
• Virtual Clusters
• Virtual Networks
Object Oriented System (OOS) or Object Oriented Design (OOD) can be implanted in the following:
•
|
Virtual Server Class
|
•
|
Virtual Container Class - contains 2 or more Virtual Sub-Servers Class
|
•
|
Virtual Cluster Class - contains 2 or more Virtual Sub-Servers + 0 or more Virtual Containers
|
•
|
Virtual Network Class - contains 0 or more Sub-Servers, Container and Clusters – can also be null
|
Advantage of Virtualization:
Network virtualization supports the complex requirements in multi-tenancy environments. Network
virtualization can deliver a virtual network within a virtual environment with independent
network resources. These virtual network can disperse traffic into zones or containers to
ensure traffic balance and resources distribution. Customizing virtual servers and networks
to meet the business requirement is the way to go. With Object oriented Design such
virtual structured environment would be composed of manageable objects and our
OO Virtual Model
would be Easy to:
• Structure and size
• Build
• Copy
• Clone
• Modify
• Test
• Inherit
• Move around
Also such virtualization can be created as a running prototype, or we can build
software programs as model virtualizer similar to ones that are used in building
airplanes, big buildings or high-rises. Such a program(s) can be created as way
to give estimates, be part of the planning as well as test other criteria without
building anything.
Virtual Networks:
A Network is all the running hardware, software, interfaces, wiring, IP addresses, licenses
and anything any network requires. In term of Network and Network Management, a network is a handful
of hardware, software, interfaces and man-hours which also has the potential of expanding and
growing out of control. To convert a network into a class or an object we need to split Virtual
Networking into sub-classes:
• Bare-Metal Sub-Class – one of more Bare-Metal Container (actual hardware as the class Attributes)
• Management Sub-Class – one of more Service manager Container
• Documentation Sub-Class – one of more Documentation Container
• Virtual Network Sub-Class – one of more Server Containers
• Exception Sub-Class – Zero of more error tracking and handling Container
• Change Control Sub-Class – Zero of more Change Control Container
• OS Container
What is a Virtual Network Class?
A Virtual Network Class (object) is an independently running network and contains all virtual
sub-classes listed.
A Virtual Network Class can be extended-inherited and reused by other Virtual Network Class.
Virtual and Bare-Metal:
To clear our virtual concept, every hardware including physical servers, switches, wiring,
etc are represented into two forms or ways as:
• Software Sub-server containing the hardware
• Properties of virtual classes
For example, the firewall hardware or box would be represented or assigned a software Virtual Sub-Server which would be the link and the handler of such hardware. Even if the hardware may be directly handling the traffic, the software Virtual Sub-Server is the representation of the firewall hardware within the Virtual Container as follows:
• The Firewall Virtual Sub-Server’s Methods will have the steps and processes
dealing, running and maintaining the firewall hardware or box.
• The Firewall Virtual Sub-Server’s Properties will be the documentation for
tracking the wiring, links, and IP addresses and so on.
• Bare-metal Container would have all the firewall physical boxes and data
including the hardware inventory and expiration tracking
Technical Notes:
As for properties, these can be implemented as a Java ArrayList<String> data type Object which
throws an exception when any of these Properties are deleted if such properties are shared
between Virtual Containers or networks. For example, routers and switches can be stored or
assigned as Properties of a Sub-Server object. They can be also shared between
Sub-Servers, containers or even virtual networks.
Network and Network Managed Services:
The breakdown of the Network and the Management Services and into objects:
Virtualization and Object Oriented Design Classes
Network Engineering and Design
High Availability
Security
Plant Design
Installation
Sub-Servers and Containers and their properties and methods
Network Architecture
Firewalls, routers and switches
Server Network Interface Card (NIC) Teaming (High Availability Architecture)
Load Balancers and Intrusion Detection Systems
WAN/LAN Integration and Storage Area Networks (SAN)
Telco Services, including Multi-Protocol Label Switching (MPLS)
Leverage Multiprotocol Label Switching (MPLS) enhanced features and functionality to effectively
Voice over Internet Protocol (VoIP)
Installation and Configuration Services
Services Management Container
Skilled Staff
Network Utilization Capacity Management
Performance and availability management, including eliminating bottlenecks
Pro-active preventive management and highly efficient reactive capabilities for unexpected repair problems
Contingency plans
Documentation Sub-Server and Container
User account administration, network devices and configuration, and change management
Standard monthly reports, including network services reports and incident reports
Exception Container
Connectivity troubleshooting and monitoring
Automated preventive notification
Pre-failure alerting
Trend analysis
Event correlation and root cause analysis
Troubleshooting and Problem Resolution
Change Request Container
Rigorous change and configuration management, including auto-discovery process and performance thresholds
OS Container
Network Operations Monitoring, alerting and reporting
iPhone Operating System (IOS)
Firmware Patch Management
Troubleshooting and Problem Resolution
Support DIR’s customers
Bare-Metal Container
All the Physical hardware
Hardware inventory
Hardware expiration tracking
Reality Check:
Using Object oriented Design and Virtualization in Building Data Centers is
new and needs all parties involved to be open-minded, think in these
terms (OOD and virtualization) and see the advantages of using these
approaches. From experience we may have an uphill battle convincing all
the parties involved.
Virtual Topology:
The key ingredient is our virtual network is the following Containers:
• Services Managers Container
• Documentation Container
• Exception Container
• Change Control Container
These Containers run the show with Services Managers Container as the Kernel or
the Boss. They should be built as a Virtual Cluster. The Second key ingredient is
the size of these classes or object should be manageable by human with automation
and intelligence. Since these Containers or clusters are virtual so creating
an array of these cluster would be manageable and can be copied, moved and
inherited. We can actually name the Center Structure cluster as “Control-Tracking
Cluster” Class. The same thing can be apply to virtual network in term of size,
management and complexity.
• Star Topology
• Mesh Topology
• Ring Topology
Data Center Specification Network Editor:
With same approach we are building cluster editor, we also developing
an editor for Network. The difference between these two editors is
the network will be using the cluster's data structure-classes and
arrays of classes to build the network classes. As for the management
and documentation, the Network editor management and documentation
classes would be the level higher in the building structure hierarchy.
Parallel Arrays of Virtual Network and Control-Tracking Cluster:
These can be viewed as parallel arrays first the Virtual Networks array and second
is the Control-Tracking Cluster array.
An admin GUI interface should be developed as the human control center for all the virtual
networks and their supportive Bare-Metal containers. The best answer to our virtual
model and approach would be best illustrated with the following topologies:
Addressing Security with Virtualization:
When it comes to security, Virtualization is a very powerful tool in creating and managing layers of security such as:
1. Virtual Proxy Servers as security front
2. Virtual Firewalls - actual protection and filters
3. Virtual Sandboxing to isolate a number of running software
4. Using Virtualization to create clusters with different functionality - adds a layer of abstraction
5. Isolation and multi-tenancy
6. Virtual Honeypots and Honeynets - decoy and traps
7. Rollback
Honeypots are computers set up for the purpose of luring attackers, and a honeynet
is an entire network of honeypots.
The honeynet looks to the outsider like a production network.
|
|
|