Download - The Modern Data Center Topology
1
The Modern Data Center Topology:
The High Availability Mantra
2
GreenField Software
• Company– GreenField Software is a privately held, early stage Indian (Kolkata-based)
software company looking to be a globally recognized player in Cloud-based Intelligent Infrastructure Management
• Mission– GFS delivers pioneering Cloud-based Intelligent Infrastructure Management
solutions to improve operational and energy efficiencies, safety and environmental conditions of facilities with critical infrastructure.
• Vision– Our Cloud-based Intelligent Infrastructure Management solutions help our
customers to• Optimize capex, reduce operating costs and mitigate risks of critical infrastructure
failures• Improve Sustainability through improved energy management and safety of their
employees and other stakeholders using the facilities
3
Partners & Customers
Oil &Gas
Media House
Telecom
Higher Education
Power Utility
Financial Services
4
Today’s Topics
• The Modern Data Center Overview
• The High Availability (HA) Mantra
• Operating Challenges
• A Solution
5
Modern Data Center Overview
6
Multiple Classes of Data Centers
• Internet Data Center used by external clients connecting from the Internet supports servers and devices required for B2C transaction-based applications (e-
commerce).
• Extranet Data Center provides support and services for external B2B partner transactions. accessed over secure VPN connections or private WAN links between the partner
network and the enterprise extranet.
• Intranet Data Center hosts applications and services mostly accessed by internal employees with
connectivity to the internal enterprise network.ness services.• Special Purpose Data Center
For specialized application areas like Geological & Geophysical for Oil & Gas Industry
May or may not be inter-connected
7
Common Objective: Business Continuity
• Disaster Recovery Data Center Each Class may have dedicated or Shared DR Center Usually located separately from Primary Data Center
• High Availability (HA) Data Center Each Data Center provided for with significant redundancies DR Center comes into play only when a Disaster strikes. Component or system failures within any DC should be either self-healing or
redundancies within the DC should take over
• Insurance Against Power & Network Outages Reliability through multiple service providers Internal Back-ups
ness services.• Securing the Data Center
Against malicious hacking that can bring down the Data Center impacting business continuity
Implementing Firewalls/ Virtual Firewalls
8
Common Complexity: Multitude of Assets
Multitude of Assets Divided between
two worlds: IT & Facilities
Includes Mission Critical Applications
Like a manufacturing operation
Raw Material: Power & Networks
Processing: Data
Output: Information Service
Needs: Asset Management, Resource Optimization, a la Manufacturing
9
The High Availability Mantra
10
Extreme Redundancies for 99.99% Uptime -> Higher Power Consumption
Huge Population of N+1/N+2 Equipment -> Asset Under utilization & Too complex tomanage with spreadsheets & Visio tools
Chain of inter-dependent equipment -> Multiple points of failures
Growing Heat Loads, Carbon Emissions & e-waste -> Sustainability Issues
KW per Rack increases as more processing capacity is added -> Trade-offs: need to support more per rack versus extra space & heat loads.
High Availability is Inversely Proportional to Asset Utilization & Energy Efficiency
Today’s High Availability Data Center
11
When HA fails - Tale of Two Disasters
Amazon RBSTech fault at RBS and Natwest
freezes millions of UK bank balances
RBS and Natwest have failed to register inbound payments for up to three days, customers have reported, leaving people unable to pay for bills, travel and even food. The banks - both owned by RBS Group - have confirmed that technical glitches have left bank accounts displaying the wrong balances and certain services unavailable. There is no fix date available.
Amazon cloud outage takes down Netflix,
Instagram, Pinterest, & more
With the critical Amazon outage, which is the second this month, we wouldn’t be surprised if these popular services started looking at other options, including Rackspace, SoftLayer, Microsoft’s Azure, and Google’s just-introduced Compute Engine. Some of Amazon’s biggest EC2 outages occurred in April and August of last year.
Which Will Be The Next One?
12
What’s the High Availability Mantra?
Amazon Data Centers (built to Tier 4 standards and with an expected availability of 99.995%) has had two outages already in 2012 – each over 3 hours!
• Tier 3/Tier 4 just defined by hardware redundancies
• Glaring gaps in operating procedures to prevent fatal human errors
• Lack of purpose-built BCP software to predict failures
• Lack of chain of custody to detect root cause
Availability % Downtime per year Downtime per month* Downtime per week 99% ("two nines") 3.65 days 7.20 hours 1.68 hours 99.5% 1.83 days 3.60 hours 50.4 minutes 99.8% 17.52 hours 86.23 minutes 20.16 minutes 99.9% ("three nines") 8.76 hours 43.8 minutes 10.1 minutes 99.95% 4.38 hours 21.56 minutes 5.04 minutes 99.99% ("four nines") 52.56 minutes 4.32 minutes 1.01 minutes
99.999% ("five nines") 5.26 minutes 25.9 seconds 6.05 seconds
99.9999% ("six nines") 31.5 seconds 2.59 seconds 0.605 seconds
99.99999% ("seven nines") 3.15 seconds 0.259 seconds 0.0605 seconds
13
Delivering the High Availability Promise
Adequate Redundancies
• Are there any points of failure – besides power and external networks - that can impact uptime? (Not everything is N+1)
• What are my redundancy paths?
• Are the relationships & dependencies among critical assets clearly defined?
• Can I do an impact analysis on the outage/downtime of any equipment? Can I predict the cascading effect of such an outage on other assets/applications in the data center?
Preventing Failures
• Can any failure be predicted to take proactive measures? Do I get alerts on threshold breaches so that I can take preventive actions before a failure happens?
• Is there a history of a Move-Add-Change (MAC) that I should be aware of?
• What is the impact of a MAC on space, power, cooling?
• Where can new devices/servers be best placed? Floor -> Rack -> Cage. How this can be determined based on current infrastructure and other dependencies to avoid a failure?
• How do I prevent a fatal human error?
14
Operating Challenges
15
The High Availability Challenge
Asset Over Provisioning Lack of HA Management Tool IT assets tracked by Systems
Management Tool
Facilities assets tracked by BMS
Two not inter-operable: Unable to determine missing link for HA
Unable to track redundancy paths
HA fails if any equipment or software in critical path fails
HA fails if there’s fatal human error
Health and history of equipment, or previous MAC impact, not tracked
Too many assets; two classes of assets
Absence of Software Portfolio (even if hardware assets are tracked)
Move-Add-Change: Decisions not based on simulations, analysis
Absence of change management
Absence of workflow approvals
Unable to predict failures
No chain of custody
Need to Predict Failures
16
Beyond HA: Infrastructure & Operational Challenges
Energy Problems Operational Problems Low level asset tracking
Under utilization of many computing resources
Running of old inefficient equipment
Decisions not based on analysis
Cooling not optimized
Floor & Rack Space: Non-optimal placements of equipment
Increasing demand for rack space
Absence of capacity planning
Higher power consumption & growing power bills
Not monitoring power use at device levels
Dissemination of enormous heat
Creation of hot spots
Drastic reduction in expected life of computing equipment
Failing of a data center
Increase in CO2 emission
17
A Solution
18
Solution That Bridges the Gap Between IT & Facilities
Data Center Infrastructure Management (DCIM) Software
IT System Performance
Management
Building Management
System
Data Center Infrastructure Management
19
Solution That Addresses The High Availability Challenge
DCIM Helps to Predict Failures
Asset Over Provisioning Lack of HA Management Tool IT assets tracked by Systems
Management Tool
Facilities assets tracked by BMS
Two not inter-operable: Unable to determine missing link for HA
Unable to track redundancy paths
HA fails if any equipment or software in critical path fails
HA fails if there’s fatal human error
Health and history of equipment, or previous MAC impact, not tracked
Too many assets; two classes of assets
Absence of Software Portfolio (even if hardware assets are tracked)
Move-Add-Change: Decisions not based on simulations, analysis
Absence of change management
Absence of workflow approvals
Unable to predict failures
No chain of custody
20
Solution That Addresses Infra & Operational Challenges
DCIM Improves Energy & Operational Efficiencies
Energy Problems Operational Problems Low level asset tracking
Under utilization of many computing resources
Running of old inefficient equipment
Decisions not based on analysis
Cooling not optimized
Floor & Rack Space: Non-optimal placements of equipment
Increasing demand for rack space
Absence of capacity planning
Higher power consumption & growing power bills
Not monitoring power use at device levels
Dissemination of enormous heat
Creation of hot spots
Drastic reduction in expected life of computing equipment
Failing of a data center
Increase in CO2 emission
21
Anatomy of a DCIM Software: GFS Crane
23
See also:Data Center Infrastructure Management: ERP for the
Data Center Manager