It’s no secret that enterprise storage requirements have exploded in the past decade,and are likely to continue to grow into the foreseeable future. The International DataCorporation predicts 30% annual growth in data storage purchasing, a fact drivenin part by digitization, big data, and increasing migration to SaaS and PaaS models.

Furthermore, 100% uptime has become a critical need to serve a market that demandsthe ability to log in and place orders instantaneously.For enterprise IT leaders, storage requirements growth translates to additionalcomplexity and escalating costs. For many, dedicated hosting is a reasonablesolution, providing greater reliability, built-in disaster planning and recovery,greater security, and better cost control. Decision makers are rightly concerned,however, about the cost and risk of making the wrong choices for their companies.

By utilizing a knowledgable and experienced hosting provider, not only are youoffsetting the capital expense of the equipment but you’re obtaining a morereliable product solution because of your reliance on the provider.

Caronet’s principals have been serving the IT industry with storage solutions since1995. We’ve helped hundreds of companies evolve their IT infrastructure to be morescalable and efficient.

We’ve leveraged the lessons of 20 years into this guide,which examines the question of whether dedicatedhosting is right for your organization, provides a stepby-step process for choosing the right provider, andhighlights the common pitfalls to avoid.

What is a Dedicated Server?

A dedicated server is a server in which you have full administrative control, butwhich the hosting provider owns. Since you do not need to purchase the server, theupfront investment is significantly less than with colocation or an in-house datacenter. Also, because the hosting company is responsible for the hardware and thenetwork, fewer resources and expertise are required for upkeep.

Leasing a dedicated server is the best option for companies who are ready for theirown server but not yet prepared to purchase one. If you’re not sure what you need,a dedicated server is a good place to start.

Critical Concerns When Choosing Dedicated Server

When choosing a dedicated server provider, consider the following factors.

Name Brand vs. White Box Servers

Reliability. Name brand servers have a higher level of reliability. White boxesgenerally use lower quality memory DIMMs and cheaper hard drives. White boxcooling fans are also known to fail sooner and the build quality on the chassis is of alower quality than name brand servers. This means that while white boxes mayappear to be a more cost effiicient decision in the beginning, the long-term cost ofownership is greater than name brand servers.

More advanced features. Dell offers a number of features unavailable on whiteboxes, such as the DRAC (Dell Remote Access Controller) for remote users toadministrate the system. With a separate physical network connection, you alwayshave access to the systems. Other system applications include diagnostic tools thatcommunicate with components such as the RAID card, so you can monitor thehealth of the individual drives of the array, the fan speeds, the voltages, the powersupply status, etc. With white boxes, RAID cards function haphazardly, voltagereadings may only be available for certain parts, and there are half as many fans dueto the driver being initially written for a very broad base of support.

Compatibility. Name brand provides a greater level of operating systemcompatibility. Generally, white box manufacturers will only test on Windowsoperating system across the board and will not test on Linux operating system.

Name brand manufacturers, for example Dell, insist that both Windows and Linuxoperating systems are tested to ensure a higher level of compatibility.

Service. The supplying vendor will often provide a higher level of service. Whennew software updates become available, you will have better service from thehigher end brand, as they will work with you to make sure that you have the latestversions.

The AMD (another option of a server CPU) market share is smaller in the server area,and Intel has the vast majority of the market. Intel is more enterprise-based andprovides a wider array of advanced features, as well as greater technical support.

Greater intergenerational support. With name brand, you can run all the same operating systems on different levels of generation hardware. This allows you tomake sure that you can grow your footprint with your business while having aprovider that is capable of having greater integration support.

Intel Vs AMD

The AMD (another option of a server CPU) market share is smaller in the server area,and Intel has the vast majority of the market. Intel is more enterprise-based andprovides a wider array of advanced features, as well as greater technical support.

Linux Vs Window

The best way to determine the optimal operating system for your companyis to evaluate your applications. Ask the following questions:

• What do you need the server for?

• What functions is it performing?

• What applications can satisfy that function?

• What operating systems does that application support?

What operating systems does your staff understand and know how to

manage?

Server CPUs vs. Desktop

Keep in mind what you are using your system for. The more resource intensive yourapplication is the greater the chance that higher end hardware will be necessary tosupport your system. While desktop and server CPUs have similar components,desktop CPUs are not built to perform at the same level as server CPUs. Server CPUsuse higher quality components, generally have more cores, can support largeramounts of memory, and perform their calculations faster. Server CPUs also offeryou greater inter-compatibility with other server grade hardware and use ECC (ErrorChecking & Corrections) memory that is fault tolerant and resistent to transienterrors. This can be extremely important to your company to ensure that critical datais not lost or corrupted.

Hardware RAID Vs Software RAID

Software RAID is controlled by the operating system. However, software RAID has anumber of vulnerabilities that make it less effective and less safe than hardwareRAID. Consider the following:

• Performance differences are dramatically more visible on heavily used hardware.

• Having the operating system involved in RAID operation increases risk.

• Hardware RAID is more fault tolerant.

IPML vs. iDRAC/iLO

iLO is the HP equivalent of Dell iDRAC. IPMl often shares the primary server

connection, so if you are having a DOS attack or you accidentally change the settingson the server and can no longer access the server, it can affect the IPMlconfigurations and you will have no way to administrate it. All of the DRAC’s andiLO’s have separate management interfaces which are generally run on entirelyseparate networks, guaranteeing that you can always manage the system if youneed to. The greater features in iDRAC and iLO provide a more polished product.

When choosing a power supply unit, consider the following questions:

Single or Dual PSU (A+B Power)

• Does the chassis support 2 redundant PSU’s?

• Does your application need 2 PSU’s?

• If you are providing redundancy, are you also providing and delivering A/B powerto the system?

When implementing true A/B power, there are two completely independent powerpaths. Both paths have at least one of the following, utility power source, maindistribution panel, back-up generator, automatic transfer switch, uninterruptiblepower supply and maintenance bypass panel, power distribution unit, rack level

PDU, and the server’s internal power supply. Each path is capable of supporting100% of the entire peak load, resulting in full redundancy and no single point offailure that can interrupt the operation of the data center equipment.

Memory Limits

When considering limits ask the following questions:

• What is the maximum amount of memory the chassis will allow you to install?

• Will the CPU and chassis support your long-term plan of growth?

• How much do you need now and how much might you need in the future?

CPU Core #’s

Properly gauging your resource requirements can either save you a lot of money bynot over purchasing or save you a lot of heartache by purchasing what you needinstead of under spec’ing in an unwise attempt to save money.

Server Variety

Can the provider offer the whole spectrum of servers, from low-end to high-end?When first starting out, you can begin with an entry level system, but as your site orapplication needs grow, your server needs will grow with it. Can your provider offeryou the ability to easily scale your server deployment? Is their a variety of servers?Can you mix high-end systems with low-end systems?

Hybrid Solutions

Cloud hosting and dedicated servers both have their benefits. Luckily, you don’tnecessarily need to choose one over the other. Working with a provider who offersboth allows you to tailor your solution to your needs, which can include both cloudand dedicated solutions.

Hybrid Solutions

For instance, it may make sense to have a dedicated server for your production area,while using the cloud for development. Because employees are only working part ofthe day, having development in the cloud means you can shut your developmentdown at night when it won’t be used. Since you’re billed in the cloud on a meteredbasis, you’ll be billed less. However, your production area is on 24/7, and aguaranteed level of performance without disruption is critical, so having a dedicatedserver will acheive those benchmarks better than the cloud.

Some providers may provide solutions that combine the benefits of each. Caronet’sDedicated 2.0 solution offers the reliability of dedicated servers with the benefits ofthe cloud.

With cross-platform system upgrades and downgrades, Dedicated 2.0 allows you tomaximize your system’s agility and seamlessly migrate from system to system as yourworkload increases or decreases, without reconfigurations. When hardware fails,

Dedicated 2.0 allows you to migrate between systems without any downtime, thanksto IaaS HA. Dedicated 2.0 also provides a number of other unique benefits, such astemplated installs/reinstalls and system cloning to accelerate new serverdeployments, 10Gb connectivity on Intel E5 servers, storage snapshots, failure/maintenance reductions, and a free, unmetered high-speed private network.

Also, working with a provider who offers cloud, colocation, and dedicated solutionsmakes it easier to transition between them or combine them as you need. The bestsolution is always one tailored to your specific situation, and you may move betweendifferent hosting solutions as your business changes.

Choosing Your Provider: Steps and Critical Factors

Once you’ve determined that a dedicated server is an important part of the solutionfor your organization, the next step is to select a good provider. Many organizationschoose based on brand recognition or marketing promises, but this can be a costlymistake. Dedicated servers are complex and their benefits and costs are impacted bya wide variety of factors. Here is what you need to know to make the right choice.

Step One: Asses Your Needs

To find a provider who offers a good solution for your company. It’s important to firstunderstand exactly what your requirements will be.

A common mistake is to significantly under-spec hardware to keep costs low. Whilethis can save money in the short term, it causes problems when you have anunexpected spike in usuage, and as your application grows. Unplanned downtime,the natural result of under-spec’d hardware, can be a game-changer, and not in agood way.

Also, if you do not have the administrative experience to operate the system securelyand to guarantee a high level of uptime, be sure your provider can help you withthese. It does not pay to unnecessarily cut costs on security and uptime.

In addition to the security and uptime, consider these factors to define your

organization’s needs:

Server management. Most providers offer the choice of managed or unmanagedservers. Do you have the staff to effectively manage the server? And if so, is it worth itfor them to spend their time on that rather than internal growth and tasks?

Software and scalability. How will your software function? How many concurrentusers will there be, what will the resource cost be, and what will your datarequirements on a per user basis be over time? The need for data is going to changeover time, so plan ahead.

Disaster planning and recovery. Data loss can be crippling for an organization at anystage, and is usually caused by either physical disaster, data corruption, or user error.

For organizations that do not yet have a budget for a complete disaster planning andrecovery system, there are lower-cost and flexible approaches to protect your datawith creative solutions tailored to your needs. A good provider can work with you tocreate such a solution.

Security. How critical is security relative to your data? What will your securityrequirements be? A dedicated server environment is often far more secure than an inhousedata center, but it’s critical to know exactly what you’re paying for both in termsof physical and virtual security.Fault tolerance. When it comes to fault tolerance, be sure to have at least RAID level

1. Raid (redundant array inexpensive disks or redundant array of independent disks) isa hardware-based level of redundancy for hard drives. In trying to be cost effective, acompany may forgo fault tolerance, but the cost of a RAID is peanuts compared to thedevastating cost of data loss.

Step Two: Assemble a list of providers

With colocation, in many cases your team provides physical management of theservers and therefore requires a data center within a reasonable distance. Withdedicated, physical management of the servers is handled by the provider, soproximity is less important than quality.

Be sure to expand your list beyond the big brand names in data centers. Often, thelarger companies are unable to provide the level of service and flexibility of packagingthat a smaller, high quality organization can offer.

Still, it’s usually a good idea to select a location that is in reasonable proximity to yourusers, to reduce the possibility of latency. If your main audience is in the US, any USlocation should be effective. If you have some users in Europe, the East Coast tends tobe better, whereas if you are serving the Pacific Rim, the West Coast tends to be best.

However, unless you are targeting a very geographically specific audience, latencyissues tend to be minimal.Be sure to expand your list beyond the big brand names in data centers. Often, thelarger companies are unable to provide the level of service and flexibility of packagingthat a smaller, high quality organization can offer.

Step Three: Evaluate providers

With a good list of dedicated providers in hand, it’s time to start narrowing them. Youcan easily rule out any that don’t meet your intitial criteria, such as security measures inplace, power and network redundancy, and uptime guarantees. With your short listidentified, it’s time to evaluate each option. Here’s your checklist.

History of Reliability

Server downtime can be extremely costly. According to a recent Poneman Institutereport, an unplanned data center outage costs an average of $7,900 per minute.

Reliable up-time depends on three main factors: power continuity, coolingeffectiveness, and network connectivity. Ask your provider how strong they are inthese areas, but don’t simply trust their stated uptime record. Expect them to havethorough RFO (reason for outage) reports, and ask them to share that data with you.

Then fact-check their records. There are a number of sites, includingWebHostingTalk.com and the Outages and NANOG mailing lists, where you cansearch for complaints and discussions on a given company’s recent outages, andcheck them against the provided RFO reports. These sites are also a great place tocheck a company’s reputation with its users.

A history of frequent outages, multiple online complaints, or falsified information inRFO reports are all major red flags. A great provider will demonstrate extremely lowoutage frequency, prompt response to any incidents, and at least 99.999% uptime.

Doing your homework on a company’s reliability is particularly important when buyingdedicated, because unlike with colocation, there are many smaller dedicated resellerswith little experience. Do your due diligence and look at a company’s history andreputation.

Physical Security

Protecting against data breaches is critical for most enterprise organizations, andoften the biggest challenge lies in physically securing the facilty where the data isstored. Security is especially important when buying dedicated because you areusually not accessing the server directly. As with reliability, it is worth doing someextra research on a provider’s capability in this area.

Make sure your provider has the latest security measures, including:

Physical Security

• Multi-layer verification process

• Multi-factor authentication

• Video surveillance

A strong multi-layer verification process should include locked external doors, lockedinternal doors, and locked cabinets. How are the identities of authorized personnelverified? Look for authentication that includes at least two factors, such as a physicalkey fob, key codes, and a form of biometric authentication such as palm recognition.

The premises should be monitored 24/7 by video and staff, to interrupt anyunauthorized access that may occur despite security protocol.

When security is a priority, finding a provider with a dedicated-only facility is thebest option. A center that shares space with colocation will experience a significantamount of traffic, creating opportunities for weakness in security. When access to theentire data center is restricted, the chances of a breach are significantly reduced.

Infrastructure Redundancy

Don’t settle for a blanket statement regarding redundancy. Due diligence inthis sphere will pay dividends when disaster strikes. You want to see redundancy onfour key elements:

• Power: Ask for full A/B power. If redundantsources of power share a path at any pointfrom the curb to the equipment, then it is atrisk of failure. A/B power means that thereare two fully parallel power paths from thecurb to the equipment, with no overlap.

• Cooling: If cooling fails, you may not know ituntil your server overheats. Look for aprovider that maintains at least twodifferent types of cooling infrastructure,such as rooftop units plus externalcondensers and CRAC units.

• Internal network: Inquire regarding the quality of the gear used to connectyour servers to the Internet, and request redundant paths as well.

• External network: Look for a provider that has peering arrangements with multiple providers to ensure a robust connection from the facility to the worldwide Internet.

Other Considerations

Security and reliability of your server is important, and so is your experience asa customer. A good provider will offer all of the following.

• Current generation hardware. Every year, the newest generation of

hardware is faster and has newer features. Look out for small houses that areonly selling access to older equipment. When a provider has the newest andmost capable equipment, you get more for the same price. A new product willlast longer, so you don’t need to do migrations and upgrades for three or fouryears.

• Support for your software. Whether you’re using Linux or Windows, or

FreeBSD, make sure your provider offers support for the software you plan torun. Some providers are Linux or Windows only.

Knowledgable deployment assistance. You may not know what the resourcerequirements are going to be yet, and it’s easy to under spec or over spec.Under spec’ing can cause disruptions and lead to downtime; over spec’ingleads to unnecessary costs. A good provider can help you determine the rightamount of end production need.

• Helpful account tools. Industry standard account tools include an onlineportal to access invoices, services purchased, bandwidth usage data, and toopen support tickets.

What to Look for in a SLA

A dedicated provider that meets all of your criteria will probably also offer a servicelevel agreement, but not all SLA’s are created equal. Read it carefully and insist on onethat:

Clearly states the uptime guarantee, with at least 5 nines of availability(99.999%).

• Backs the guarantee with significant financial impact to the provider should theguarantee not be met. You should expect a reasonable refund or credit in theevent of outage, not merely a few dollars a day.

• If the SLA guarantees 100% uptime but the provider does not have A/B power andredundant networking, then they cannot meet their guarantee: Look elsewhere.

The SLA should represent an organization that is not afraid to back its guaranteewith real consequences.

The Relationship

In the course of this detailed investigation of dedicated providers, you will naturallyget a sense of the priorities and approach of each provider. Does the provider seem

reluctant to build a true relationship? Do they have you speaking to a differentsalesperson every time you call? If either of these are true, how responsive will they bewhen you need support?

When buying dedicated, the relationship is especially important, since your provider willbe maintaining the servers. Look for a provider that demonstrates a clear interest inbuilding a relationship. While this may seem like a minor point on the face of it, considerthat when an issue arises, these are the people you will be relying on to keep yourbusiness running.

Avoiding Common Pitfalls

Pitfall #1: Bundled Pricing Structure

When buying a dedicated server, make sure you know what you are paying for. Somecompanies bundle management into their pricing, and while the support may be great, ifyou’re not utilizing it, you’re overpaying for the services you’re actually using. If yourcompany has the capacity to manage the server yourself, opt for an unmanaged dedicated server, and ask to unbundle any other services you don’t expect to use.

Pitfall #2: Poor Quality of Support

When problems arise, your provider’s support team should be able to answer yourquestions and solve your issues, but quality of support varies among providers. Youdon’t want to wait until you have an issue to find out whether your provider is one of thegood ones. Check an organization’s online reputation, and carefully assess the level ofservice they provide during ramp-up to determine whether they will serve your supportneeds.

Pitfall #3: Inability to Scale Technical Requirements

The provider may not have current generation hardware or hardware that can expandand support the technical requirements of your company. If your application is growingand you need greater performance, can your provider offer sixteen core servers if youneed them, or do they only provide 8 or 12? With growing storage needs, can yourprovider offer terabyte after terabyte at a performance metric that’s right for you?

Our product offerings include Dell dedicated servers, private cages, cabinets, half-cabinetsand/or Rackspace by the Unit and our Cloud offering. The data centers have UPS battery backup power and multiple diesel generators onsite. Transit includes several Tier-1 backbone providers which are directly accessed using multiple dedicated fiber routes. With over 60gbps of dedicated bandwidth, Caro-net is one of the largest data centers in Southeastern United States.