Dedicated Server Purchasing Guide

Why do you need a dedicated server?

Before you commit yourself to a dedicated hosting environment, it might be worth checking that you really do need it. We see three core reasons for putting customers on to dedicated servers:

  1. Performance
  2. Control
  3. Security

Performance

Performance can be the motivator for going dedicated for two reasons. The application to be hosted has grown to the point that it generates more load than is reasonable on a shared web hosting platform. Alternatively, the nature of the application being hosted is such that performance is critical: under no circumstances can a drop in responsiveness be tolerated, and the budget is available to ensure this is the case.

Control

A shared web hosting environment by definition means that as an end user, the amount of control that you can have over the server and its configuration has to be restricted. Root or Administrator level access obviously cannot be provided. While your hosting company will usually work as hard as they can to accommodate your needs, some requests simply wont be possible within a shared environment.

Unless you are on a dedicated server, you surrender a degree of control of the server to the hosting company. This has the advantages of ensuring all changes are made by professional Systems Administrators but there are the disadvantages that if you are a budding Admin yourself, you have to wait for someone else to make changes on your behalf.

There are many possible server configurations that will conflict with the sometimes restrictive environment that exists in shared hosting. On a dedicated platform you have full control over how the server is configured since the changes are of no consequence to other users.

Security

While shared servers are locked down to very high levels, the very fact that the server is accessible to other authorised users does introduce increased security risks in respect to host compromises. On a dedicated server you can be sure that any security breaches will only be a result of your own actions and not of somebody else.

Hardware specifications

One of the more obvious decisions is deciding on the hardware requirements for your server. We recommend selecting hardware that you think will meet your requirements for the first 12 months as a minimum. If you don't allow for this, the cost of upgrading the hardware can often more than exceed the costs of deploying the higher specification hardware in the first place.

Every application is different and so the loads that it will place on a server are very difficult to estimate. Pages, hits, visitors, data transfer/month can all act as a guide but the variability in load from one application to another can be so significant (and implementation-specific) that these figures might not tell you much.

Hardware components

Let's have a look at the core hardware components in your server so you can see which ones warrant consideration when specifying a server. This table only attempts to discuss the components in the context of your ability to upgrade them in future, and the effect of the failure of this component on the delivery of services.

Component

Physical/hardware factors

Operating system factors

Failure factors

Power supply

Chose between Single or Dual (redundant). Upgrading usually requires a chassis change.

Linux: Chassis upgrade to gain dual power supply doesn't require operating system reinstallation.
Windows: Most cases requires complete operating system reinstallation and application configuration required for upgrade, if the motherboard chipset has changed.

Power supply failure will take your server offline, but generally doesn't cause loss of data.
Many data centres only guarantee power to one power supply OR the other, not both. Dual power supply is required to make use of the power supply uptime guarantees.

CPU

Upgrading requires only a 15 minute period of scheduled downtime. The amount of performance increase will be limited by the server chassis.

No effect on the operating system

Failure will take the server offline, but generally doesn't cause loss or corruption of data.

Memory - Capacity upgrades

Upgrading requires only a 15 minute period of scheduled downtime. The amount of capacity increase available will be limited by the server chassis. Higher end server generally accommodate more RAM.

No effect on the operating system

Will take the server offline, but generally doesn't cause loss of data.

Memory - standard vs ECC

As for Capacity upgrades. In addition, the server motherboard must have ECC support. Generally found in higher-end servers.

No effect on the operating system

ECC memory provides a level of protection against data corruption in machines running constantly, for example, servers. An essential component to any server.

Hard disc - Hotswap vs Internal

Hotswap drive bays tend to be used on higher-end server equipment. These permit rapid removal and replacement of hard discs.

Hotswap systems allow drive changes/replacements without shutting down the server. Internal drives require the server to be shut down.

Elimination of downtime during replacements of failed hard drives. Minimises downtime during live capacity upgrades.

Hard disc - SATA vs SAS

SAS provides more I/O capacity than SATA. Typically requires higher-end server hardware.

Applications that are more I/O intensive (e.g. database servers) will benefit more from SAS drives. SATA drives offer greater and more cost effective storage capacities.
Upgrade from SATA to SAS is possible if supported by chassis.

No appreciable difference.

Hard disc - Capacity upgrades

Different chassis have capacity for varying numbers of drives bays, typically 2 to 8. For large storage requirements, the chassis may dictate the total possible storage capacity.

Live capacity upgrades can be done under Linux. They are not possible on Windows 2003. Capacity upgrades are both risky and time consuming and should be avoided where possible by allowing for 1-2 years growth requirements at the time of server provisioning.

N/A

Hard disc - RAID configuration

The chassis must have capacity for the number of drives for the chosen RAID configuration.

Each different RAID configuration has different performance and cost factors. See our RAID configurations article for more detail.

A minimum of RAID1 is recommended to protect data during drive failure. RAID0 or single drive systems should not be used in servers.

You should also consider the pro's and con's of load balanced and clustered server configurations in considering hardware specifications.

Hardware brand names

Different dedicated hosting companies will have their own hardware preferences ranging from brand names such as IBM, Dell, HP or Sun to lesser but still good quality such as Supermicro, or even white-label hardware.

The question is - should this matter to you? In a managed hardware service - it is your hosting provider that is responsible for ensuring that the hardware is working correctly.

Bandwidth

Most dedicated servers come with an initial data transfer allowance. This can vary anywhere from 5 Gigabytes in Australia through to Terabytes on offer by some overseas hosting providers.

Before you get carried away with choosing the host with the most bandwidth, try to work out how much bandwidth you actually need, then make sure the hosting package you're choosing meets that requirement.

Providers that offer excessively high bandwidth allocations may not always be the best choice. They may also be attractive to other high bandwidth clients that subsequently cause congestion on the network, or relate to a service that is not providing the same quality of bandwidth as other providers.

In comparing the bandwidth allocation, our article on dedicated server bandwidth billing models may also be of assistance.

Reporting

While the server specifications are an easy component to identify in the selection process, there are other important things to make sure you get with your managed server.

Data reporting

Most dedicated servers will come with a base data usage allocation. Anything exceeding that, however, will attract excess usage charges.

It's important that you do monitor your usage to make sure it's in line with your business expectations, and that you have the ability to easily monitor it.

You should ensure that either regular emailed reports are available or more ideally a web-based interface is provided to let you login and query data usage at any time.

Availability monitoring

Depending on the service you purchase, your dedicated hosting provider will to varying degrees be responsible for monitoring the availability of your service.

If you need to know when services are going up or down (stop being available or become available again) it is important that your service provider is able to provide you with these reports. You should check for email or SMS based alert options.

Support

Support is by far the most difficult component of a dedicated server package to accurately compare when choosing between hosts.

We like to consider the provision of managed hosting services as a combination of three components:

  1. Hardware: the actual server hardware provisioned for you - as discussed above.
  2. The Environment: consisting of the Data Centre, the network and ancillary shared services that go into keeping the hosted services online.
  3. Support: the ongoing management of the software on your server.

For most users, it's the support services that will constitute the majority of the interaction with the hosting company after a server is provisioned. Making sure the appropriate level of support is both available and provisioned for you is what will ensure that your application remains online.

The reason that most people use a managed hosting service is that they have an application that they need to be kept online 24 x 7, with as close to 100% uptime as possible. This goal cannot be achieved without a lot of work, the same way that you can't expect a car to always run without fault without proper routine servicing. The higher the demands that are placed on the service, the more maintenance that will be required. To continue the analogy - consider the significant difference in maintenance of your family sedan with a race car.

If you're buying a managed server - it's important to make sure that you're not just getting hardware, but also the level of support that is right for you. If you don't, you can be assured that it will either turn up in the form of consultancy fees after you've signed the contract, or worse yet, you'll suffer from outages as a result of the much needed maintenance not being carried out.

We often think of the support that is needed to maintain uptime as consisting of three core components:

  • Prevention
  • Detection, and
  • Response

The more preventative work that is undertaken and systems that are monitored, the greater the chances of picking up changes to a service before they result in outages.

When considering alternative dedicated hosting providers, find out exactly what level of support is included in the quoted monthly charges and what services will attract additional fees.

Depending on your level of skills or interest in performing systems administration tasks, you need to make sure that the right level of service is being provided to keep the server online.

Typical support and maintenance tasks that you should ask about are:

  • Application and security updates
    • Will operating system and application updates and security patches be applied to my server?
    • Do these relate to those from the operating system vendor or any applications you have on the server?
    • How frequently will they be applied?
    • What time of day will they be applied? Do I have a choice?
  • Monitoring
    • What services will be monitored?
    • How frequently does the monitoring occur?
    • What level of reporting is available?
    • What form of response is provided to problems when they are detected?
    • What happens if my website isn't working but the monitoring doesn't detect a problem?
  • Firewall & Security

    • Is a firewall configured on the server or on a dedicated firewall device?
    • Who is responsible for maintaining it and making configuration changes?
    • Are there any systems in place to detect if the security of my server gets compromised?
  • Application installation
    • What applications will be installed at the time the server is built?
    • What happens if I need additional applications installed at a later date?
  • Configuration management
    • Who is responsible for making configuration changes to the server?
    • Do changes cost each time or are they included in the monthly fees?
  • Performance
    • If my application is running slowly will you help to diagnose the problem?
    • Do you collect any data on server load trends over time?
  • Provision of support services
    • Can I call and have questions answered over the phone or will I have to email all support requests?
    • Will I be able to talk directly with the support staff that build and maintain my server?
    • What is the typical turn around time on non-urgent/urgent support requests?

Ideally your hosting provider will be able to offer you with a support pack that covers either the rudimentary tasks required to keep your server secure, or better yet, the services that are needed to meet your expectations of support.

Any such support pack that is offered for a fixed monthly fee will have a defined scope to avoid it costing the hosting company too much money in the case of the more demanding clients. Find out what the scope is, or what the limits are if you're choosing a support pack. Make sure they are documented as part of the contract.

Additional support services

On day one you may only need a simple hosting service, but as your demands grow, make sure the hosting company has the capacity to grow with you, make sure they can provide services such as:

  • Backups
  • Monitoring
  • VPN
  • SSL
  • IDS
  • Managed firewalls
  • Load balancing

Who is the hosting company?

There are some quick checks you can do on the dedicated hosting companies you are considering purchasing services from:

  • Is dedicated hosting core part of their business?
    • This is important to ensure that there is focus on this service.
    • It ensures that the benefits of ongoing product development will flow through to you.
    • It means you can expect a consistent and reliable level of service.
  • How many staff do they have? or what is the size of the company?
    • With the 24 x 7 demands of hosting operations a very small team may be able to provide very personal service but it may not be possible to always be available when something goes wrong.
    • We discussed hardware redundancy at length above, redundancy in people is also important, if any of the systems administrators are on annual leave, is there someone with sufficient technical skills to solve problems in the event of outages?
    • There can also be trappings with very large companies. The reputation may offer some comfort but keep in mind the pro's and con's of what may become a less personal service.
  • What happens after hours?
    • Find out how outages are dealt with outside of office hours.
    • Do they have staff on shifts, in the NOC 24 x 7?
    • Are high level technical staff available to solve problems that first level 24 x 7 staff may not be able to?
  • How do they deal with hardware failures
    • Does the company rely on on-site warranties provided by the hardware manufacturer? If so, find out what guarantees there are on replacement times.
    • Preferably the company will adopt a standardised hardware platform and maintain its own inventory of spare components to cover any hardware failures.
    • Remember that a commitment of 99.8% uptime represents around 1.5 hours of downtime per month, a 4 hour turn around on replacement of failed components will cause significantly more downtime than an on-site spares inventory.
  • Find out about network management
    • One of the more dynamic parts of the hosting service is the network connectivity. It's subject not just to the performance of the upstream providers but also the activities of all of the hosting providers customers and the people that interact with their hosted services.
    • Ensure the provider has redundant (more than 1) links with different providers.
    • Ensure there is good connectivity to other domestic carriers as this is likely to be where most of your clients are.
    • Ask for a network diagram to get more information and certainty about the network configuration.
    • Find out if they manage the network directly themselves or is it outsourced to another company? If it's directly controlled, it means that the hosting provider can constantly develop and improve their systems, react to any recurring faults. It means there won't be as much passing the blame in the event of outages.

Pricing

While you'll discover all of the prices that relate to the services you initially purchase, it's worth getting a feel for what the optional extras may cost you down the track. Cover points such as:

  • Hardware upgrades - RAM, CPU, and Hard discs
  • After hours support if needed
  • Hourly rate for consulting work
  • Costs for configuration changes

Good luck!