Archive for the ‘Data Center Uptime’ Category
The illusion of data center uptime
Most of the mid-size companies that visit our Midwest colocation facility already have a data center. It’s the one in their home office. These companies have built a data center inside the four walls to take advantage of real estate that is already leased, along with cheap, fast network access for all of the employees in the building.
Redundant generators protect against data center downtime
Some companies value data center uptime more than others. These companies are in markets where their computer downtime can cost them sales, profits and clients. These companies often have in-house data centers with more sophisticated equipment to keep the computer systems up and running in the event of a power outage. These companies invest tens of thousands of capital dollars in battery backup, power conditioning and generators to protect from downtime. A few even spend thousands more in capital dollars to makes the air conditioning more reliable.
But do all these data center capital costs improve uptime? The answer is yes, but in many cases, not enough. Many of us mistakenly look at the last five years of actual downtime to judge whether our data center is highly reliable. This is a mistake. Your data center may not be reliable, even though you’ve been lucky for the last five years.
What does it take to keep your downtime to less that an hour per year? It takes data center with two of everything that is critical for operation: power, cooling, and communications systems. This two of everything model is also called N+N data center redundancy. Without it, companies should expect hours or days of downtime per year.
Uptime Institute uses a structured system to classify data centers. Tier IV data centers are built with N+N redundancy (two of everything) to maximize reliability. These Tier IV data centers are designed to deliver 99.995% uptime, which is 28 minutes of downtime per year or less. But building a Tier IV data center is expensive. A second power feed into a building can cost a quarter of a million dollars. CFOs routinely reject the idea a second generator because of the exorbitant capital costs. Without N+N data center redundancy, the uptime numbers just don’t add up.
What’s the answer to high uptime and manageable costs? Many companies use affordable wholesale colocation facilities. Some of these outsource data centers offer 99.995% uptime in exchange for monthly operating expenses rather than exorbitant capital costs. Many IT staffers use colocation to reduce their workload, get out of the power and cooling business, and focusing their data center management on their critical computer systems.
Colocation is not for every company. Applications, users, geography and other factors play into whether colocation or cloud computing might improve the reliability of your data center. The bottom line is the cost of downtime to your company. If you need 99.995% uptime, don’t fall prey to the illusion of data center uptime. Consider wholesale colocation to solve the uptime problem and manage data center costs.
Why do hardened data center facilities matter? Hardened data centers reduce the risk of a prolonged outage due to natural and man-made disasters.
On the west coast, data centers are hardened by a number of different technologies to withstand the frequent shaking of earthquakes. In the Midwest, hardened data center facilities protect against natural disasters such as tornadoes. F5 tornado resistant data centers protect your mission critical facilities and equipment from natural disasters.
Although the risk of a natural disaster is low, the risk of a prolonged outage in a natural disaster is very high.
Is your enterprise data center protected by a hardened data center building?
Consider how Cloud Computing’s unique characteristics will change how you do architecture
Today is a wonderful time for anyone interested in Cloud Computing to be working with the US government. On the one hand, the government considers Cloud to be strategically important, and they already have a track record as an early adopter of Cloud Computing on a grand scale. On the other hand, the government is also in the unique position of being able to drive standards for the approach—and in fact, they are even responsible for establishing the most widely adopted definition of Cloud Computing.
The federal agency who has taken this leadership position is the National Institute for Standards and Technology (NIST), an agency of the US Department of Commerce. NIST’s formal definition of Cloud Computing is already well known—“a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.” Concise as that definition is, it only marks the beginning of the work NIST is doing to formalize and standardize the full breadth of Cloud Computing approaches, both within the government as well as for the world at large.
More of the Cloud Computing Journal article from Jason Bloomberg
Data center building, power, and cooling disciplines are not IT disciplines.
Your expertise on applications, software architecture, network, server and storage design is not expertise on building tier IV data centers with 99.995% uptime.
Likewise, experts on mission critical facilities like hardened data center buildings, data center power redundancy and cooling are rarely experts on mission critical systems and applications.
A best-of-breed CIO strategy would include expertise in both information technology systems design and highly available data center facilities. How is this done?
If your organization likes to “roll your own” enterprise data center, you probably hire design/build experts to help you accomplish your goals of high data center uptime. Although the capital costs associated with in-house data centers can be enormous, internal data centers offer the highest level of control.
If your organization is considering outsourcing the facilities disciplines, wholesale colocation offers a simple way to offload the “landlord” side of the data center without losing control of the systems.
It’s often best to outsource data center facilities when you’re great at IT but not so great at building data centers.
Midwest colocation facilities like Lifeline Data Centers offer F5 tornado resistant buildings,N+N power and cooling redundancy, and access to many telecom providers. Midwest data centers offer low power costs also give you peace of mind that you’ve done the best job at solving the data center downtime problem using an affordable colocation solution.
Are you trying to be an expert in both facilities and IT? Talk it over with the mission critical facilities experts.
Should your outsourced data center (colocation) provider also be your IT services provider?
Your purchasing department would probably say yes. Your legal department might too. One throat to choke. They’re looking at the problem from a vendor management perspective, and fewer vendors is better. Or is it?
Ask yourself these questions:
What if you love the data center facility but the quality of the IT services offered are marginal?
What if you already have vendors for specific IT services?
What if you prefer to choose best of breed vendors for specific projects and technical support?
What if your staff does most of the IT work?
Would it be more sensible to separate the choice of data center provider from the choice of IT services provider?
Most companies that choose wholesale data centers over self-built data centers make the decision based on the uptime they get per dollar spent. That’s because these pure data centers, also known as wholesale colocation, concentrate on one thing: mission critical facilities. 99.995% uptime requires incredible attention to detail with hardened data center buildings, redundant power, cooling, telecom access, and data center regulatory compliance. But not all colocation providers are alike; data center reliability varies greatly based on the companies power, cooling, telecom systems, and compliance. If data center uptime is important, then the sensible approach would be to pick the best-designed facility for your needs.
Does it make sense to reject the best-fit data center facilities provider because they don’t do router work, or AS/400 support, or eCommerce platform support? The answer could be yes. It depends on your organization’s applications and your own staff’s talent in supporting these business-specific applications and their platforms. When considering full-service providers, make sure that you understand the quality of the data center behind the provider’s services. You have the option to pick your own wholesale colocation facility and require your vendors to support the hardware in your colocation space.
If you’re purchasing rack space from a full-service provider, you may be paying too much for your colocation space. Especially if your provider maintains a large staff of IT Support Engineers. Bench time is expensive, and unless these Engineers are fully utilized, your rack space pays for part of the Engineers’ wages. Make sure you consider competitive pricing from other colocation facilities. Data center pricing models are excellent indicators of what vendors value and how they handle their overhead.
Your outsource data center provider does not have to be your IT services provider. You have options. You can choose the best among data center vendors with a little homework.
Do you need a hardened data center? If you are trying to protect your data center’s ability to withstand natural or man-made disasters and acts of terrorism, then the answer is yes. The term “hardened data center” described computer room facilities that are designed to withstand tornadoes, earthquakes, and man-made disasters.
Many data center strategies includes using hardened data centers to achieve 99.995% uptime. Hardened data centers, along with redundant data center power and cooling all play a part on maximing reliability and increasing data center uptime.
Not all hardened data centers are alike. Many believe the best type is a reinforced concrete structure. Reinforced concrete offers the best protection against tornadoes, the most common natural disaster in Midwest data centers and Midwest colocation. F5 tornado resistant data centers help companies protect against wost case. Reinforced concrete also offers excellent protection from earthquakes.
Some hardened data centers are constructed as a building within a building, based on the thinking that a natural disaster might destroy the outermost building while the inner building protects the data center. This approach is often used when an existing building is being refit as a data center facility. The effectiveness of the protection is completely dependent on the type of building and construction.
Hardened data centers can be used as primary (production) data center facilities or as disaster recovery sites. Many consider a hardened facility more important for the production environment in order to minimize service interruptions.
How important is a hardened data center? It is critically important if you are trying to avoid downtime and if your area is prone to disaster.
Do your worst-behaved applications define your data center requirements?
In Part 1, I talked about the most important applications in your business, also know as your mission critical applications. I covered reasons that many of these mission critical applications are ill-behaved and require special care and feeding in your enterprise data center. These reasons include high bandwidth requirements on the headquarters or wide area network, expensive overbuilt servers, and additional hours of maintenance overhead per month.
How do these worst-behaved applications affect your data center requirements?
Bandwidth – Many applications generate large amounts of network traffic for even the smallest user activity. These levels of traffic can cause the applications to perform poorly via the Internet or small remote office connections. These apps can influence your decision on the location of the primary data center. It often seems simplest to place the data center close (in the same building) to your highest number of users. Yet an in-house data center may not fully support your data center uptime requirements.
Interoperability – If your most important applications link up to other important applications, you may be forced to put these applications in the same data center. If your manufacturing system is feeding data to your customer management system and your accounting system, reliability becomes more important, because a small amount downtime can affect three important software systems, not just one.
Souped-up, expensive servers – Experience has taught your IT staff to overbuild server and storage hardware to solve some of the bad behaviors of your mission critical applications. These non-standard configurations can drive up costs. Non-standard configurations are also more difficult to operate in cloud computing environments, forcing the data center to remain physical, instead of virtual.
More Maintenance – More problems mean more maintenance work to solve them. This drives up FTE requirements and makes outsourcing more complex and expensive. Maintenance load can influence location and staff requirements for the data center.
Costly uptime – Problem applications are harder to keep running and often require more technology for high uptime levels. Expensive high uptime technologies like clustering greatly drive up the costs of keeping the application alive and well.
Ill-behaved line-of-business applications influence strategic data center decisions:
Primary data center location – Would your data center be better off in-house, in the cloud, in an outsourced data center facility or a hybrid of all three?
Wide area network design – Where is the hub of the network? How many telecom providers should I use? How much bandwidth do I buy? How can I get the best pricing?
Server hardware ownership and maintenance – Do I buy my own servers for maximum control? Do I use virtual servers in the cloud? Do I use a combination of both?
Maintenance – Does in-house staff do maintenance or do I outsource it?
Good CIO strategy includes a clear understanding of the mission critical applications and their data center requirements.
More CIOs are using these tools to mitigate the risks of their worst-behaved applications:
- Thin application delivery via software by VMware and Citrix to solve bandwidth problems
- Affordable colocation to build in 99.995% uptime on the data center power and cooling
- Cloud computing services like virtual private servers for predictable mission critical applications
- Change management discipline to manage application behaviors and reduce maintenance
Don’t let your worst-behaved applications cause you to make bad decisions about your data center.
Do your worst-behaved applications define your data center requirements? If so, you’re not alone. Your most important applications are often the hardest applications for IT to support and to keep running. These mission-critical applications and their behaviors greatly influence design of your data center.
What are some examples of mission critical applications?
- A manufacturer’s most important application might be the production line management system.
- A sales rep organization’s most important application is probably the Customer Relationship Management (CRM) system.
- A lawn service organization’s most important application might be the home-grown application they paid to have developed that allows them to track clients, contracts, on-site visits and route scheduling.
- A large financial services firm may have many “most important” applications, such as client/account management, real-time updates from the markets, statement printing and archiving.
Why are some of your applications ill-behaved?
Ill-behaved applications are applications that require unusually high levels of server, network, and/or maintenance requirements. Industry specific applications are notorious for being ill-behaved. Problems include input/output intensive processing, server and workstation memory hogging, and network traffic flooding.
Application problems are often related to the software source:
- Your organization may have paid for a team to develop a company specific system. That team may not fully understand the resource requirements of the software they develop until it is finished and in use.
- A small software company with expertise in your industry may have developed the application for a single client, then sold their application to other organizations in the same industry. This small software company’s expertise in software development directly affects the behavior of the application.
- A Large software companies may have been developing applications for your industry for years. They revise the application as technology changes, but the application is largely the same application written 10 or 20 years prior. This old code base often has resource requirements outside of today’s “normal” server and network configurations.
Your IT staff knows if you have ill-behaved applications. These applications are more difficult for the IT staff to maintain. Yet the organization requires that these applications always be available. 99.995% uptime seems to be the expectation. Data center uptime is now critical, even in small organizations.
In part 2, I’ll cover how these ill-behaved applications affect data center design.
RIM did more dancing around the issues than frank sharing as it tried to explain the BlackBerry outage–leaving CIOs to speculate. And goodwill’s running short.
After an almost four-day outage of RIM’s Blackberry service, RIM’s co-CEOs gave a status update Thursday morning. Mike Lazaridis delivered what appeared to be a prepared statement, followed by questions, largely from the media. The way that RIM reacted to the outage will likely shape the company’s fortunes for the foreseeable future. And on the key question, the future health of RIM’s network and its ability to scale, too many questions went unanswered.
Lazaridis started out with an apology and something of a promise. “You expect better of us, I expect better of us,” he said. “We are, and will take every action feasibly, to minimize the risk of this happening again.”
Apparently, one switch’s failure with a bonked-up backup system had such a tremendous “ripple effect” that it caused a world-wide outage for days.
The question that many CIOs and CTOs are asking is, if architecture is planned out right, and testing occurred on a reasonably diligent basis, how exactly could that happen?
More of the Information Week article from Jonathan Feldman
Data Center Knowledge recently conducted its 2011 survey of enterprise class data center operators and found that data center design is still the top area of interest. The survey has be conducted in each of the last four years, and in 2010 data center design moved to the top of the list and remains a key topic of interest for 2011.
Having the proper design infrastructure eases the challenges associated with operations and scalability. Data center operators are charged with simultaneously saving money and making their facilities more sustainable to comply with corporate social responsibility (CSR) policies. Design is the key tool that can incorporate both objectives and deliver results.
Power and cooling has moved up the list from number three to the second position on the list. Power is often the number one cost to operate the data center so it’s no surprise this is an area of strong interest. Cooling projects are often the easiest way for data center operators to realize short-term savings in optimizing their data centers, so decision makers are keenly focused on identifying the best cooling solutions.
More of the Data Center Knowledge article from Kevin Normandeau
