The following presentation was given during the Spring 2012 Data Center World conference in Las Vegas, NV by Kirk Killian. Learn more about Data Center World at www.datacenterworld.com.
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Understanding And Evaluating Colocation Data Centers
1. Interested in data center management?
Learn about the data center management sessions offered at the
upcoming Fall 2012 Data Center World Conference at:
www.datacenterworld.com.
This presentation was given during the Spring, 2012 Data Center World Conference and Expo.
Contents contained are owned by AFCOM and Data Center World and can only be reused with
the express permission of ACOM. Questions or for permission contact: jater@afcom.com.
2. Understanding & Evaluating
Colocation Data Centers
Kirk A. Killian, Executive Vice President
214/365-2050 (direct)
214/244-1111 (mobile)
E-Mail: kkillian@pnmcf.com
10000 N. Central Expressway, Suite 1001
Dallas, TX 75231
2
3. Classifications: Enterprise Colo,
Individual Cabinets & Web Hosting
Enterprise Colo: Suites/Cages with some non-shared
infrastructure (PDU’s, CRACs, UPS, generators, electrical
switchgears, chiller plants) and non-shared telecom circuits
Colo Cabinets/Racks: Lease of one or a group of cabinets or
racks with fully shared critical systems infrastructure
Web Hosting: Provider owns servers and all critical
systems infrastructures and user uploads applications and data
This presentation focuses on Enterprise Colo.
3
4. Key Colo Terminology
Provider – the company operating the colocation facility
Pod – a phase or major section of a data center, usually with
self-contained critical systems components
Suite – the premises occupied solely by one customer, often
with hard walls and non-shared infrastructure components
Cage – same space as “Suite” occupied by an enterprise
customer, but less likely to have dedicated (non-shared) PDUs,
CRACs and UPS modules.
Critical Load – the IT equipment load, typically backed up by
UPS and generator.
Modular – design with critical systems components
manufactured off site and assembled at the data center.
4
5. More Colo Terminology
Managed Services – services often provided by a colo company
in addition to the basics of space, power and telecom circuits;
May include server re-boots, rack and stack, server builds, tape
changes, tape shipping/receiving, SAN/NAS usage, firewall, load
balancing, and intrusion detection.
Carrier Neutral – colo customer selects whichever 3rd party
telecom carriers it chooses to serve its suite.
Home Run Circuit, in telecom, a circuit provided directly by a
telecom company to an end-user, usually through or around a
colo facility “Meet Me” Room.
Meet Me Room – room within a colo facility where telecom
providers install their circuits into the facility, from which user
circuits are provisioned to colo customers via cross connects.
5
6. History – Colocation’s
Early Days
Rapid growth in 1998 – 2000
Mostly TUI Tier 1 & Tier 2 Facilities; A Few Tier 3
Mostly 50-100 watts/SF and 12 – 24” raised floor
Large open data halls w/ cages & cabinets for lessees
Limited acceptance by traditional corporate IT departments
Ready acceptance by internet companies, where speed to
market has been paramount and few internet companies had a
deep bench of legacy data center facility expertise.
Industry shakeout in 2001 – 2004 with bankruptcies,
consolidations and closures: The “Dot Bomb” era.
6
7. Modern Colo:
2006 to Present
Steady, strong growth since 2006
Newer facilities with higher reliability - mostly TUI Tier 3 and
“almost Tier 3” infrastructures with substantial concurrent
maintainability
Higher densities, mostly 100 – 300 watts/SF; some much higher
More experienced providers
More sophisticated providers, many with SAS70/SSAE16 audits,
PCI compliance, and professional operations and reporting
Financially stronger providers, including publicly-traded choices
Cost reductions per kW of Critical Load
More dedicated (non-shared) infrastructure than in earlier era
7
8. Modern Colo:
Dedicated Infrastructure
“Pod” data center deployments with less infrastructure sharing:
Key concept for some corporate customers as suites for 300 –
2,000 kW critical loads not likely to share PDUs, CRACs, UPS
modules, perhaps even chiller plants, primary switchgears or
generators with other customers (“data center within a data
center” concept.)
Example: Colo suite provides 400 kW Critical Load to (1) customer.
Dual-feed PDUs and CRACs non-shared within suite.
N+1 UPS modules (3 x 300 kVa) non-shared serving just this
suite.
Primary switchgear, generators, chiller plant shared with (2)
other corporate users in this data center pod.
8
9. Modern Colo: Gradual
Corporate IT Acceptance
Changes in colo industry have resulted in increased corporate
acceptance of Enterprise Colo model since 2006.
Example Corporate Commentary:
1999: “I don’t do colo.”
2003: “My company’s not ready for colo yet.”
2007: “I might do colo; but not on this project.”
2010: “How quickly and cheaply can I be live? My
timeline is short and my CapEx budget was cut.”
9
10. Types of Colo Providers
Large National/International Provider
Regional Provider
Local “Mom and Pop” Provider
Wholesale Only
10
11. Large National/
International Providers
Many data center sites, offering potential for geographic
dispersion
Full suite of optional managed services
Experienced team with deep bench for problem solving
Financial strength, ability to back up SLAs
Capital markets access for future infrastructure expansion
funding
Less flexible rules & policies than smaller, entrepreneurial
providers
Less responsive than smaller, entrepreneurial providers
More expensive than other options, some much more
11
12. Regional Providers
Many benefits of national/international providers
More flexible & responsive than largest providers
Less geographic reach than largest providers, but great local
expertise
Less financial transparency than largest providers
Mix of older legacy facilities and new state-of-the-art data
centers
Usually less financial strength than largest providers, but still
stable, dependable entities
12
13. Local “Mom and Pop”
Providers
Entrepreneurial, nimble, very responsive
Passionate about providing great customer service
Customer is “Big Fish/Small Pond”
Operations team may be inexperienced
Shallow labor bench for problem solving
Minimal financial strength for expansions or backing up SLAs
May have renovated and expanded an older data center on a
shoestring
Unlikely to have SAS70 Type II certification, PCI compliance,
and sophisticated operations & maintenance procedures
including audit trails
Being purchased and merged into larger companies rapidly
Poor corporate acceptance into this segment
13
14. Wholesale Providers
Pre-designed “Pods” and critical systems infrastructure,
increasingly designed with better power scalability
Lowest cost/kW, but inflexible pricing for growth
Limited managed services
Longer lease commitments required (6-15 years)
Newer facilities (<= 5 years old) with reliable infrastructure
May not have SAS70/SSAE16 certification, but very experienced
working with outside auditors in getting one
Strong financial condition and financial capacity to construct
expansion space for growth
Wholesale providers can build a suite to your specifications, but
this requires design, permitting, construction & commissioning,
typically requiring 5-8 months.
14
15. Colo Trend #1:
Higher Electrical Densities
New colo facilities have electrical densities of 100 – 300
watts/SF.
Some new colo facilities have initial electrical densities of
100-175 watts/SF, but are designed for future increases to
200-300 watts/SF.
Some facilities claim ability to cool 1,500 watts/SF, using full hot
or cold aisle containment, ultra-sophisticated and expensive
multi-path cooling systems, with 300 watts/SF becoming readily
attainable w/o containment systems.
15
16. Colo Trend #2:
Modular Designs
Several colo providers are offering modular designs, where they
offer “data center in a box” solutions, allowing the customer to
custom design and implement new data center pods as needed.
Shipping containers with cabinets, “power packs” and cooling.
Offers advantage of paying for some capacity only when needed,
saving both CapEx and OpEx.
Offers perceived advantage of less critical systems sharing.
16
17. Colo Trend #3:
Cloud Services
Colo providers are offering an ever wider array of optional
“cloud” services, including SAN/NAS, “burst capacity” servers,
network services, security services, applications management,
etc.
Optional managed services allow the small user to elevate their
game without ramping up staff.
Many services offered 100% ala carte, separate from other colo
contract requirements, valuable as the customer gets to know
the provider better over time.
17
18. Example Enterprise Colo
Pricing Models
Most common pricing models are:
D. Cost/kW of Capacity + Metered Power
F. Cost/SF + Electrical Circuits
18
19. A. $/kW of Capacity + Metered Power
This pricing model includes:
1) monthly cost/kW of installed capacity (not actual usage)
2) metered power, plus
3) 20-70% uplift on metered power for HVAC & common area
electricity loads
Customer pays a one-time charge (“NRC”) for electrical circuit
installation, usually without provider markup (negotiable)
If you use significantly less power than your required capacity
you’re overpaying for the provider’s infrastructure
Very efficient if you can predict actual usage accurately
Great pricing model if provider can add critical systems
components over your contract term to increase your capacity
as needed.
Some providers charging on actual usage instead of capacity.
19
20. B. $/SF +Electrical Circuit Charges
This pricing model includes:
1) monthly cost/SF for your suite space, plus
2) monthly charge for specific electrical circuits installed; actual
electricity usage is usually included in the circuit charge
Individual electrical circuits are expensive but “pay as you go”,
so you can have them installed as needed, provided you’ve
reserved the capacity from your provider.
Can be cost effective if growth is non-linear and unpredictable.
Be sure to reserve growth capacity, or provider will sell your
future growth power/cooling to other customers – ouch!
Reserve “B” – backup circuits (for dual corded servers) often
priced at 50% of “A” - active circuits.
20
21. Is Enterprise Colo
Right For You?
Good Fit:
Low Capital Expenditure Budget
Immediate Need
Short-Term Need
Physical Need for IT Equipment in a Non-Core Geography
3rd Party Managed Services Need
Power/Space Quantities Well Defined for 3-5 Years
Huge Carrier Diversity Needed
Poor Fit:
Lowest Long-Term Occupancy Costs Required
Maximum Flexibility Required in Changing Critical Systems
Infrastructure
Maximum Control Over Critical Systems Operations Required
21
22. The Baker’s Dozen: Tips for
Evaluating & Selecting a Colo Provider
(13) Recommendations in colo evaluation and selection
22
23. 1. Carefully Define Your
Provider/Facility Criteria
Carefully define your colo criteria at the project outset:
Geography
Systemic Redundancy/Reliability
Minimum Space Required
Power Requirements By Year (Define Early in Project!)
Growth in Power (tough to estimate accurately – see next slide)
Telecom
Facility Hazard/Risk Tolerance
Desired Managed Services
Required Financial Strength/History of Provider
Project Timing
Project Budget
23
24. 2. Estimate Growth in
Power & Cooling
Define anticipated power needs by year (not easy; especially
after year 3)
Verify each provider’s systemic ability to provide additional
power/cooling, including physical placement of additional
infrastructure.
If growth is sought in power, verify the provider’s ability to
remove the additional heat (It’s generally easier to increase
power than cooling.)
Get a mechanical engineer’s opinion of future cooling design if
you’re doubtful (i.e. 200 watts/SF using 12” raised floor).
Ask about highest density area in facility and talk to that
customer about hotspots & provider responsiveness.
Request a CFD analysis to prove cooling capabilities.
Structuring the colo contract to accommodate uncertain growth
but not overpay for that growth is the key challenge.
24
25. 3. Seek Carrier
Neutrality
“Carrier Neutrality” means the colo provider lets you select your
telecom providers without provider surcharges, offering:
More lit carriers (typically)
Lower pricing by encouraging competitive pricing among carriers
Higher likelihood of getting truly redundant fiber networks
A dark fiber option, which is great if you are operating
active/active mirrored or DR data centers.
25
26. 4. Prepare RFP to Get
Competitive Proposals
Prepare Requests For Proposal to get competitive proposals.
Send RFPs to 4-8 wisely selected providers; research candidates
best suited to meet your needs and be respectful of providers’
proposal prep time.
Be realistic about quantity of information sought in proposals.
RFPs encourage providers to submit competitive proposals.
RFP should clearly communicate your project needs,
expectations and timing.
An outside advisor/consultant can assist with the RFP process if
you’re not experienced in colo RFPs.
26
27. 5. Analyze Proposals and
Negotiate Contract Terms
Input your space/power growth needs into each provider’s
pricing model.
Compare proposals under multiple growth scenarios
(anticipated, high-growth, low-growth, etc.).
Pricing comparison will help match provider’s infrastructure
scalability against your growth needs.
Seek line item pricing, and compare pricing elements across
providers.
Many elements of provider proposals are negotiable.
Include telecom costs in pricing comparison; may vary among
providers, especially if they have discount carrier options.
8-40% savings are typical from well-written & strategically
executed RFP and proposal negotiation process.
27
28. 6. Evaluate Providers’
Financial Strength
Some providers are publicly-traded; some have S&P/Moody’s
credit ratings.
Get audited financial statements (after executing Non-Disclosure
Agreement).
Get financing reference if future construction is required.
Get assistance from your CFO’s office if available.
Bankruptcy or closure of a colo provider you’re under contract
with can be disastrous, so do your homework. This is very rare!
28
29. 7. Get Customer
References
For your colo provider finalists, get customer references, not just
from the providers’ 1-2 favorite customers (the easy ones), but
from their 3 largest and a few smaller ones. (Not always easy to
identify and interview.)
Ask references not only about the provider’s actual performance
and uptime, but also their responsiveness to that customer’s
evolving needs.
Ask the customer if they would structure their colo contract
differently if they could repeat the process and why.
29
30. 8. Set Realistic Uptime Expectations
Be realistic in uptime expectations: compare uptime targets
against basic systemic redundancies.
Compare SLA terms early in your proposal review.
SLA credits are usually weak compared with customers’ costs.
Providers work very hard to protect their uptime reputations.
Uptime is generally very good for newer facilities with well
executed maintenance and operations processes.
Biggest downtime problems are in older facilities that have been
inexpensively and quickly expanded and poorly maintained.
Many providers exaggerate TUI tier ratings; very few are Tier 4.
30
31. 9. Select Provider With A
Great Cultural Match
Interview provider’s key personnel – match interviews with your
team SME’s.
Is the provider’s team and corporate culture going to work well
with yours?
This is a long-term “going steady” relationship – not a prom
date.
Strive for Win/Win deal terms; Punitive Win/Lose economic
negotiation is usually a bad bargain.
A great relationship is most important when extensive managed
services or a flexible growth model is required.
31
32. 10. Form Multi-Disciplined
Colo Selection Team
Enlist SMEs from your organization: Hardware, Applications,
Electrical, Cooling, Facilities, Telecom, Security, Business
Continuity, Financial and Legal
Designate a project coordinator.
Engage an outside advisor if your coordinator hasn’t evaluated
colo before or is too busy.
Outside advisors are generally much less expensive than the
time and money savings they provide.
32
33. 11. Get Contract
Options
Seek flexibility through reasonable options in colo contracts:
Power/cooling growth option is very important.
Contraction Right should be reasonable (advance notice, return
highly re-leasable subset of cage, pay subdivision costs, etc.).
Early Termination Right should be reasonable (advance notice,
limited termination windows).
Termination Option for Poor Performance should be verifiable.
Get a Renewal Option at fixed rates, not “market” rates,
including modest increases over original pricing.
Seek a one-year Extension Option in addition to other renewal
option term.
Ability to add managed services at pre-determined prices.
Contract options are one-sided to customer’s benefit: get them!
33
34. 12. Study Future or In-
Progress Construction
Scrutinize critical dates on provider’s construction schedule.
Verify funding sources for future or in-progress construction.
95% of project steps may occur on time but the other 5% can
put construction project weeks/months behind schedule.
Assume projects not already underway will not deliver as
originally promised.
Some providers have promoted “vaporware” growth plans.
Allow one month for new data center commissioning.
Decide if you’re comfortable being the first customer in a new
data center, especially regarding telecom circuits.
34
35. 13. Allow Adequate
Project Timeline
Reasonable colo selection project timeline includes:
1 month – Establish needs
1 month – Research prospective providers
1 month – Write Request For Proposal
1 month – Get proposals & evaluate
1 month – Negotiate terms & select preferred provider(s)
1 month – Negotiate/execute contract
0-6 Months – provider builds colo suite/cage
= 6 – 12 Months for most organizations
Several steps can be completed simultaneously.
Process can be accelerated under certain circumstances.
35
36. Conclusion
Enterprise colo can be a great solution for the corporate data
center user under certain circumstances.
Colo facilities and providers are not “One Size Fits All”, so
evaluate and select wisely.
Carefully selecting and negotiating with providers can yield
significant savings and a contract with flexibility tailored to your
specific needs.
36
37. Evaluating a Colocation
Data Center
Call or E-Mail With Any Follow-Up Questions:
Kirk A. Killian, Executive Vice President
214/365-2050 (direct)
214/244-1111 (mobile)
E-Mail: kkillian@pnmcf.com
10000 N. Central Expressway, Suite 1001
Dallas, TX 75231
37
38. Interested in data center management?
Learn about the data center management sessions offered at the
upcoming Fall 2012 Data Center World Conference at:
www.datacenterworld.com.
This presentation was given during the Spring, 2012 Data Center World Conference and Expo.
Contents contained are owned by AFCOM and Data Center World and can only be reused with
the express permission of ACOM. Questions or for permission contact: jater@afcom.com.