How virtual data centers & cloud technology work

Technical Info: A Deep Dive Into ProfitBricks

How has ProfitBricks become the Cloud industry's price/performance leader?

InfiniBand, KVM, dedicated CPU and RAM, dual redundant RAID 10 storage, massively configurable private cloud networks. Take a closer look at how all of these things, and more, contribute to making ProfitBricks different from other public cloud vendors.

KVM Loves Infiniband

Under The Hood

InfiniBand & KVM Working Together

How ProfitBricks Combines Proven Technologies to Enable the Public Cloud Features You've Been Waiting For

InfiniBand has long been a staple of the supercomputing community with use in 80% of the prestigious Top 500 supercomputing clusters. Harnessing that speed for public cloud, ProfitBricks runs dual channel InfiniBand, providing an incredible 80Gbps capacity to our network compared to the 1Gbps other cloud providers use.

ProfitBricks doesn't simply settle for using that as a fast connection between virtual machines, though. We extended KVM, the widely used open source hypervisor, to take advantage of InfiniBand and enable customer-facing features like flexible virtual machine sizes (from 1-62 CPU Cores & 1 - 240GB RAM),live vertical scaling, and superior performance with dedicated CPU cores and RAM that are never shared between instances.

How did we do all that? Click the "More" button and find out.

More

ProfitBricks Infiniband

Our InfiniBand-Powered Modular Architecture

You may have noticed, most public cloud providers offer instances in stock sizes, with fixed ratios of CPU cores, RAM, and temporary storage. They do that to try make it easier on themselves to stack virtual machines in even intervals on top of their physical hosts. In the process, they make you buy more non-redundant cloud storage, most of which you cannot use unless you add your own redundancy solution, as the amount of CPU cores and RAM gets larger.

With a design unique to the public cloud industry, ProfitBricks physical hosts contain no storage. Instead, all volumes are attached across the network from the hosts to a dual-redundant RAID 10 storage array.  Since the connection occurs over our 80Gbps InfiniBand channel, disk performance exceeds many of our competitors attached, temporary storage solutions but with a redundancy that stores 4 copies of each data block.

Our extended version of KVM, without storage attached to the hosts to worry about, has the ability to move a virtual machine from one physical host to another so fast that the virtual machine doesn't know it's been moved.  How?  KVM only has to contend with the memory allocated in the instance and has an 80Gbps InfiniBand channel in which to move it.

This approach gives ProfitBricks a degree of freedom when optimizing our physical hosts that no other cloud platform has.  We don't have to force cookie-cutter instance sizes on our customers since we have the ability to move virtual machines among the hosts to more efficiently use the hardware.  This is how we are able to offer flexible instance sizes and live vertical scaling without a reboot.

The architecture is so efficient, we are able to dedicate CPU core and RAM to individual virtual machines without sharing resources, enabling performance gains not possible on other cloud platforms.

Want even more detail?  Keep reading subsequent sections of this page below…

More

Instance Size

Flexibility & Live Vertical Scaling

You choose your size, including the biggest instance sizes available anywhere if that's what you need.  Plus, add CPU cores & RAM without rebooting

When it comes to scaling, Cloud 1.0 providers force you to do so horizontally where you add more servers to a cluster. They seem to have forgotten all about vertical scaling, the ability to add CPU cores and RAM to accommodate your increased load.

Need a big cloud server?  The other guys limit you to packages in specific CPU core or RAM amounts.  At ProfitBricks, you can go up to 62 CPU cores and 240 GB of RAM, the largest available in the public cloud industry.

Find out how we utilize our unique, InfiniBand-powered, modular architecture to provide customizable instance sizes that can scale vertically without a reboot by clicking "More" below.

More

Read The White Paper On Flexibility Download

Custom Instance Sizes, Change It Without Rebooting, or Go Big

Take a look at this zoomed in screen shot from our easy-to-use graphical user interface, the Data Center Designer.  For the number of CPU cores and for the amount of RAM, those are sliders that let you choose the exact size you want.  Because we can better utilize our hardware with our ability to move VMs among different physical hosts without interrupting service, we aren't beholden to stock sizes to make our job easier like our competition is and we pass that flexibility onto our customers.

You choose from 1 to 62 CPU cores and up to 240 GB of RAM, so you can get precisely what you need instead of what you are forced to buy with some stock instance size that has too much of one resource or too little of another.  At ProfitBricks, you can max out both CPU cores and RAM to the full size of our physical hosts, 62 CPU Cores and 240 GB RAM, for the biggest instance size available in the industry.

What if your needs change?  Other public cloud providers give you one choice: scale horizontally by adding more instances to your footprint.  With ProfitBricks, not only can you horizontally scale, but vertically scale by adding CPU cores and RAM to an existing VM.  You don't even have to reboot it with the ability to scale vertically, LIVE.

An optional configuration as shown in the screenshot, ProfitBricks achieves this by signaling to the operating system via an ACPI (Advanced Configuration and Power Interface) event. Operating systems like Ubuntu, CentOS, and many others understand this signal, enabling the instance to use the new resources without a restart.

Read the Neovise White Paper explaining why flexibility matters Download

More
Cloud performance driven by great technology

Performance Matters

Dedicated CPU Cores & RAM + 80 Gbps Network

No time slicing or over provisioning of CPU or RAM.

Like all public cloud providers, ProfitBricks pools resources.  Unlike our competitors, however, once a CPU core or an amount of RAM has been allocated to a specific virtual machine, that instance gets exclusive use of those resources.  That eliminates noisy neighbor problems present on other clouds and eliminates the performance inconsistencies seen when a virtual machine has to share a physical processor with another instance.

Add to that the 80Gbps InfiniBand channel between the instances and the result is industry-leading performance.  Want to see standardized test results like UnixBench (for CPU and RAM), Dbench (for disk), and iPerf (for network)?  Download the white paper or get a preview by clicking "More" below.

More

Read Cloud Spectator's Performance Report Download

3rd Party Benchmarking: ProfitBricks vs Amazon EC2 & Rackspace Cloud

Chart showing results of UnixBench test of ProfitBricks vs. AWS and Rackspace

A preview of the Cloud Spectator benchmarking results is shown here with some UnixBench comparisons.  A widely-used standard to evaluate CPU core and RAM performance, UnixBench is actually a compilation of multiple tests.  Things like double precision math operations, shell script concurrency, and process creation times are aggregated together and given an index score.  The higher the score, the better the performance.

This graph shows two common sizes at Amazon EC2 and two more at Rackspace Cloud compared to same-sized instances on ProfitBricks.  While your individual application may impart different load on a virtual machine, UnixBench serves as a good guideline for overall speed.  Not only do the test results show ProfitBricks to be at least twice as fast as Amazon EC2 and Rackspace Cloud, you can pay less for the same performance on ProfitBricks.

For example, look at the m1.medium score for ProfitBricks compared to the m1.xlarge score on Amazon EC2.  They are nearly identical yet the 1 CPU, 3.75 GB RAM instance will be far less expensive.  Similarly, a 2 CPU, 4GB RAM machine on ProfitBricks outperforms the 6 CPU, 15 GB RAM instance on Rackspace Cloud.

Performance absolutely matters.  Not just so your applications run fast, but so that you can pay less for equivalent performance you have to pay more for elsewhere.

Read Cloud Spectator's Performance Report Download

More
How ProfitBricks stores data.

Storage

Fast, Dual Redundant RAID 10 Block Volumes

InfiniBand and tiered caching make the difference

Each ProfitBricks data center is separated into two Availability Zones, with a fire section border in between them.  Each block volume is RAID 10 in each Availability Zone with the entire set of RAID 10 data blocks instantly replicated to the other Availability Zone.

This is why, unlike the CPU cores and RAM, you cannot select which Availability Zone for volumes, because they automatically reside in both.  RAID 10, x2.

Yet another benefit of the ProfitBricks 80Gbps InfiniBand is that this additional mirroring happens over an incredibly fast channel.  Factor in the RAM and SSD caching on our storage array, you get 4 copies of each data block while still maintaining superior Dbench scores to our competition.

Cloud networking technology

Networks

Any Private SDN Imaginable

Why Limit Yourself to One Public and One Private Network?

Software Defined Networking (SDN) is a huge topic these days, but most public cloud providers limit you to placing a virtual machine on a public network or in a private one, but not both.  Each ProfitBricks virtual machine has up to 8 virtual network cards, allowing you to segment your instances any way you wish on multiple networks simultaneously.  This diagram, taken from our Data Center Designer graphical user interface, shows a relatively simple scenario involving a LAMP stack (Linux, Apache, MySQL, & PHP).

The two Apache/PHP servers are connected to the public Internet via a load balancer, utilizing one of those network cards each (shown in blue).  However, in order to limit exposure to the MySQL database, the two Apache/PHP servers make use of another of their virtual network cards to take part in a private network with the data source (shown in red).

The MySQL master database uses a second network card as well, to connect to its slave so that a continuous mirror can be kept at the database level (shown in orange).

This is just one example, but you are only limited by your imagination when segmenting your data into different private network segments, all made easy to see with our popular Data Center Designer visualizations.

Easy to use cloud technology

Easy to use

API for Devs, CLI for Sysadmins GUI for Everyone Else

Visual Presentation That Makes Sense or Complete Programmatic Freedom, You Choose

The best way to plan and architect Internet environments is to use a large, blank whiteboard to outline, correct and improve the complete infrastructure design by hand using a pen.  The relationships between the components are complex and the easiest way to keep them all straight is to represent them visually.

First generation cloud hosting IaaS providers require you to convert these drawings into tables of text meant to indicate virtual servers, storage and how they interact with one another - a cumbersome and error prone process. Plus, with each subsequent change you have to pick the right row in the table and setup network connections, IPs and more.

ProfitBricks offers three ways to easily build and manage your Cloud:

  • The revolutionary Data Center Designer, visualizing cloud assets for everyone through a Graphical User Interface (GUI)
  • Our Command Line Interface (CLI) which makes building scripts easy for system administrators
  • SOAP (HTMLPDF) and REST (coming soon) APIs for developers who want programmatic control

We eat our own dog food at ProfitBricks: the Data Center Designer is JavaScript written on top of the SOAP API and the CLI is Python built on the same.  Our developer community actively creates integrations on top of the API for popular developer tools/languages like Chef and Ruby.

Cloud security technology

Security

Baked In, Not Bolted On

The Many Security Advantages of the ProfitBricks Cloud

  • Unmatched physical data center security with US footprint co-located at SwitchNap near Las Vegas, NV
  • Secure data transmission connection via TLS (Transport Layer Security)
  • Strong industry-leading data deletion policy
  • Redundant systems guarantee fast, smooth troubleshooting with the highest level of availability across multiple fire zones
  • Current, most up-to-date, high-quality software, continuously monitored and maintained by experts
  • SSH Key Pairs for encrypted login sessions
  • Full control of port-level firewall on each instance
Cloud technology support

Support

Support: Free Phone-In, From Experienced Sysadmins

If you have a question, shouldn't you have an actual, professional human being to talk to instead of looking through public bulletin boards?

A lot of public cloud companies think customer service is optional, offering self-service or limited telephone support with the base cost of their product.  We think that's just plain wrong.

Why?  Because when you have a problem you need help and you need it right away.  When you have a technical question, who better to ask than a senior system administrator who understands your issues and can offer clear solutions directly and without delay?  By phone (+1-866-936-0764) or by email (support-us@profitbricks.com), you get to choose and it's always 100% free.

That's right, there are no hidden or additional costs for support.

At ProfitBricks, Technical Support is Free.

Prefer to take a crack at it yourself first?

Check out our FAQ and our Knowledge Base for answers to commonly encountered problems.