In the previous articles in our server buying guide series, we’ve compared the pros and cons of on-premise vs. cloud servers and provided a guide to help you figure out how best to configure a server including a deep-dive into what options are most important to consider before making a purchase. In this, the final article of our series, we go in-depth on each of the available form factors (rack, blade and tower) and help you understand the differences between the most popular manufacturers.
- What form factor is best based on your needs
- Most important differences between rack, blade and tower servers
- Manufacturers differences for product and after-purchase support
What is Server Form Factor?
The term form factor used in this article describes the physical shape of the chassis and housing of the server’s hardware. Beyond size and shape, each form factor has its own unique requirements for housing, maintenance and usage. After reading this article, you’ll be able to choose the server form factor that’s the best choice for your datacenter or business needs.
What we’ll be reviewing
||A rack, or rack-mounted server comes completely functional right out of the box and is built explicitly to be mounted into a server rack. Rack servers are high-performing and can be configured to support a wide range of requirements
||A blade server is a CPU & memory circuit that can be vertically mounted with other blade servers to form a cluster in a blade chassis enclosure. Blade servers offer an extremely high processing density, but often require shared components like storage and cooling to operate
||A tower server is an extremely popular, easy-to-configure server that can sit under a desk like a standard office PC. Tower servers have a large footprint that makes them better suited for office environments
Rack servers are a type of higher-end, easily expandable server that can be vertically mounted into physical cabinets called racks. Both the servers, and the racks that hold them, adhere to standardized sizes called rack units (abbreviated as U or RU) that measures 1.75 inches (44.45 mm) in height. Although they’re available in a variety of sizes, the most common height for a standard full rack cabinet is 42U – just over six feet – and while it’s important to keep in mind that power requirements, technician access and cable management can often lower these numbers in real-world applications, conceptually you can think of a 42U rack having the following capabilities:
- 42 1U rack servers
- 21 2U rack servers
- 10 4U rack servers
Most IT professionals who manage medium to large data centers will opt for 2U (and up) servers because the larger physical volume of the taller chassis allows for better airflow and larger, more powerful cooling fans that results in higher heat dissipation. Along with improved cooling, the larger chassis size of 2U (and up) servers means that they can accommodate more drive bays and expansion cards, which increases the overall capabilities of the server. Lastly, power consumption of higher U servers is often lower than a comparable number of single Uservers would be when mounted in a single rack.
Rack servers offer high-performance, moderate pricing and come equipped to run as a stand-alone system. Perfect for large offices, data centers & co-location facilities
What is a socket?
A socket denotes the maximum number of central processing units (CPU) a rack server can be configured with. A 1-socket server supports a single processor, a 2-socket supports up to two processors, a 4-socket supports up to four processors etc…While it’s true that dual and quad socket motherboards, like the uber popular Zeon line from Intel offer significant performance benefits over single CPU motherboards, the more recent advent of CPU’s with extremely high core counts, like the 32-core EPYC/Threadripper line from AMD, have resulted in an increased popularity of single-socket CPU’s for their overall cost-effectiveness.
Advantages of rack servers
- Fully Functional: Unlike blade servers, which work best when paired together and often require shared components like power and cooling, rack servers contain all the components necessary to function by default – including storage and memory – which makes them ideal to operate as stand-alone units or paired together data centers.
- Convenience: The ability to mount multiple servers on to a single rack makes installation and maintenance easier, and saves floor space in data centers.
- Cost-effective as stand-alone units: Although rack servers can be clustered together, they don’t have to be, which makes them much more cost effective to own and operate – even if you only need a handful of units for your desired applications.
Disadvantages of rack servers
- High Power Requirement: While overall power consumption can be reduced somewhat by utilizing 2U and higher configurations, an entire full rack of servers will require a large amount of energy to run. Furthermore, when the rack is densely populated, the racks will need more cooling units. This will raise energy costs and increase the risk of overheating when many servers are stored next to each other.
- Time Consuming: Locating the source of a problem in a rack server system can be time-consuming.
You might need to invest in a unique cooling system to prevent issues from arising.
Blade servers are a thin, modular type of server that mount vertically into chassis to save space and offer improved processing density. The more compact design and mounting arrangement that allows for the potential for higher CPU density is due to the fact that – unlike fully self-contained rack servers – most blade servers are stripped down to only containing CPUs, network controllers and memory. While most blade servers do come with some internal storage options, it’s at a very limited capacity, which means that each blade in the cluster is often set up to utilize shared storage devices that are built-in, or connected-to the chassis enclosure. All other components required for blade servers to operate and connect to the network – like ports, and power connectors and cooling – are also designed to be shared via the chassis of the blade enclosure. Lastly, it’s important to note that while blade servers are often clustered together, by default each unit is setup to operate independently of one another. That being said, IT administrators can use system software to create a blade cluster, which is software that allows a group of independent blade servers to work together as single, high processing density computing unit.
Blade servers are the most expensive, but they offer the highest processing power in the smallest footprint and are easier to manage. Perfect for large data centers.
What’s a blade enclosure & does it fit into a rack chassis?
A blade enclosure is the physical housing that holds multiple, vertically-aligned blade servers and facilitates the connection for the necessary shared components to setup a cluster such as storage, cables, networking, power and cooling. The number of blade servers that will fit into a single enclosure varies by manufacturer, and by the processor count of the server (the higher the processor count, the more width it will take in the enclosure). Enclosures are designed to fit into rack chassis with popular size variants being three rack units (3U) and six rack units (6U) tall.
An IT administrator hot-swaps a blade server from the chassis enclosure.
Advantages of blade servers
- Highest processing density: Utilization of shared components like storage, power and cooling allow for hyper-dense computing power in a relatively small footprint
- Cable and systems management: Unlike rack and tower servers that require independent configuration of storage, cables, ports, cooling, etc… each blade in a cluster is hot-swappable and utilizes the same shared components, which means single-point of access for maintenance and upgrades
- Low energy spend: Rather than having to independently power and cool multiple servers located in the same rack – as is the case with rack servers – the blade server cluster is powered and cooled through the chassis enclosure, which reduces overall energy spend substantially
Disadvantages of blade servers
- Upfront costs: New blade servers cost more to buy than a similarly configured rack or tower server (although it’s important to note that price-points of refurbished blade servers are often much less expensive). However, blades do offer long-term cost benefits from a more simplified ability to manage, lower overall energy usage and real-world performance benefits when clustered together
- Requires cluster to operate efficiently: Designed to run in parallel with other blade servers, a single blade offers little-to-no benefit over a rack server as you’ll still need to account for missing components such as additional storage, power, cooling, etc…
Less flexibility: Blade Servers are usually seen as less flexible than rack servers. It’s often not possible to equip the external RAID card for blade servers with a disk array, which would be necessary for higher-resource database applications. Furthermore, clusters of multiple blade servers are required when the application use-case requires a large amount of memory.
A tower server, which has a chassis that resembles the size and shape of a standard desktop PC, is in-fact the most common type of server. Like rack servers, tower servers are completely self-contained units and they do not share any components like storage, cooling or power. Although highly customizable, tower servers often come with minimal hardware components and pre-loaded software, as they tailer to plug-and-play administrators and the needs of small to medium-sized offices. The physical footprint of a tower server is quite large relative to the available computing power, so organizations considering this type of server need to make sure they have enough space to dedicate. Tower servers are highly customizable and they can be configured for a variety of purposes including web, network, communication or general purpose.
Who are tower servers best suited for?
Tower servers are perfect devices for small businesses, law firms or healthcare organizations that need a server that can be easily set up and managed without a dedicated team of IT administrators on staff at all times. Tower servers can be used for running internal applications, serving public websites or storing sensitive customer data. Compared to rack and blade alternatives, Tower servers are also much easier and less time consuming to customize and configure based on specific business needs.
Tower servers are the least expensive and highly configurable, but offer low processing density and the largest physical footprint. Perfect for small offices, healthcare organizations and law firms.
Advantages of tower servers
- Ease of use: Tower servers come with minimal configuration required, so small IT teams with less resources can easily customize and upgrade based on business needs
- Cost While tower servers will never have the computing power or density of racks or blades, as single, stand-alone devices – they’re often cheaper to purchase
- Low cooling cost: The large physical footprint of tower servers inherently means that they have a much lower component density than blade or rack servers. A lower component density requires much less cooling
Disadvantages of tower servers
- Large physical footprint: Tower servers aren’t designed to be neatly stacked or mounted, so if multiple servers are required to operate, the physical footprint that they take up will build quickly
- Low computing density: Tower servers can be fantastic stand-alone devices and they can be configured to be quite powerful. However, their large footprint and inability for multiple towers to cluster their computing power together make them a poor choice for larger businesses and data centers
- Peripheral management issues: In environments where multiple tower servers are in operation together, IT administrators might have to invest in additional cables, switches or even have to replug external devices into each separate server, which can quickly become messy and difficult to maintain over time
According to market research from International Data Corporation (IDC) worldwide server shipments grew 18.4% year over year to nearly 3.2 million units in the second quarter of 2020 and two manufacturers were head-and-shoulders above the rest: Hewlett Packard Enterprise (HPE) and Dell. Although both companies lost market-share compared to the same quarter of 2019, HPE and Dell accounted for a whopping 28% – 14.9% & 13.9% respectively – of all servers sold worldwide. In terms of sales revenue and units sold, HPE came out on top with $3.582 Billion and 456,642 units while Dell, who as recently as 2018 claimed top stop had $3.339 Billion / 432,556 units.
Dell vs. HPE
Both Dell and HPE have tremendously capable servers and offer a variety of options for all three of the form factors detailed in this article. As with any battle of the brands there will be loyalists on both sides. The objective in the comparisons below is to capture some of the sentiment from the community of system admins and IT professionals who work on these machines every day so that it might help you the next time you need to make a purchase.
The popular belief among many system administrators and IT professionals is that Dell servers cause the least administrative problems and have the fastest cold starts. Considering affordable prices, they are often the preferred choice both when building storage systems and when organizing solutions for providers.
Advantages of Dell servers
- Scalability: Hardware is standardized, and therefore has the ability to upgrade with the absence of compatibility problems between different models.
- Warranty and support services: Dell after-sale support is well-regarded and has good technical support and service. They provide a warranty with on-site service to last as long as 7 years, offer round-the-clock technical support and complete free documentations.
Disadvantages of Dell servers
- High cost: You have to pay a high price for high-quality servers. Nevertheless, Dell products cost less than similar equipment from other well-known manufacturers (HP, Lenovo/ IBM, Intel), but cost more than Supermicro.
- Periodic stock issues: Some people report that, at times, it can be difficult to get a Dell server configured with the exact component and features you want. The manufacturer predetermines the configuration and capabilities of commercially available servers, so you either have to pay extra for redundant functionality or purchase options additionally.
- Update issues: While virtually non-existent with the lastest generations, some older generation PowerEdge servers have had reports of issues installing firmware updates by users
HPE and their line of ProLiant server models are extremely popular, which is backed up by the company’s continued lead in market-share worldwide. This is not surprising since they are characterized by high quality, performance, and good scalability. Also, HPE servers continually improve their monitoring and management technologies.
Advantages of HPE servers
- Price flexibility: HPE is considered to be more flexible when it comes to pricing, even though the initial quotes for HPE are similar to that of dell. In some cases, your relationship with the vendor plays an important role. There are a lot of HPE used servers out there, and this makes it easy to find and get spare parts
- Management tools: HPE server management platform iLO comes standard and is included in the purchase price. The GUI and general user experience of iLO is thought to be more intuitive and faster than iDRAC – the management software for Dell servers. It should be noted that the more advanced version of the tool does require the purchase of an additional license, which can be used to lock you in with the OEM if your servers go EOL.
Disadvantages of HPE servers
- Delivery: HP servers are mostly shipped in pieces that you will need to assemble and test at your IT provider’s lab. That means it takes more time for you to set up your servers and get them tested. However, HP’s shipping time is 1-2 days, while that of dell can take up to 2 weeks.
- Reliability: Some system admins on the web have reported increased as much as 2.5x downtime with HPE hardware as compared to similar hardware from Dell. That being said, HPE’s predictive alerts are useful for managing said downtime.
- Post-sale support: HPE servers all have extremely thorough documentation, which all can be found on their website, but official after-sale support from HPE requires a contract. Some users also report difficulty in finding support for drivers and firmware updates for 2nd and 3rd generation machines
If you’re in the market for a rack, blade or tower server from Dell or HPE, we have one the largest inventories of new and refurbished servers on the market – and we sell them for thousands less than the OEM. The latest models of Dell PowerEdge and HP Proliant servers, 2-year warranty and US-based phone support on every order. Start a chat today to learn more!