The Value Of True Dedicated Computing

Filed in Cloud Industry Insights by John Engates | July 16, 2013 10:00 am

Dedicated hardware is one of the pillars of the Rackspace Hybrid Cloud[1], which blends dedicated servers with public cloud and private cloud to create the best fit for each customer’s specific IT needs. Many of our customers come to us from a one-size-fits-all public cloud, and they tell us that the hybrid cloud approach has improved the performance, reliability and overall cost of their infrastructure.

One thing we’ve learned from these customers is that dedicated hardware (whether bare metal or with virtualization) isn’t going away. If anything, there’s a renewed need for it with the increased use of I/O-heavy applications such as databases and Big Data platforms. The public cloud is a powerful technology, but it isn’t the answer for every business or every workload. Customers want to run the cloud where they want (whether on premise or in a vendor’s datacenter), how they want, and in the combination that best fits their applications. In many cases, dedicated hardware will play a key role.

We’re not the only ones discussing the importance of dedicated computing. Amazon Web Services recently chopped the price of a product it calls “EC2 dedicated instances.” As part of that move, AWS cut its “dedicated per region fee” — the extra fee charged to a customer for each region in which it runs a dedicated cloud instance — by 80 percent, from $10 an hour to $2. AWS also reduced by lesser amounts the cost of the dedicated instances themselves. But a lower unit price doesn’t always mean lower costs overall. Nor does it always deliver value when one considers an apples-to-apples comparison of performance and support.

We at Rackspace don’t aspire to offer the lowest unit prices. We strive instead to offer the best value, which includes the value of our award-winning Fanatical Support. We encourage customers to consider the total cost for the specific level of performance and support that their business requires. Our hybrid cloud approach helps each customer find the best fit — in terms of performance and cost — for each of its workloads, whether on public cloud, private cloud or dedicated hardware.

As noted by analysts at Wells Fargo and Cowen and Co.[2], Amazon’s move provides clear validation that dedicated, single-tenant computing will continue to serve as a critical component of IT infrastructure, and that some workloads run far better there than in multi-tenant environments. That message might seem to contradict repeated public statements by Amazon executives that every workload belongs in the public cloud[3]. But the real conflict lies in the way that AWS defines dedicated computing, which is at odds with the view of the rest of the industry, including Rackspace.

EC2 dedicated instances are essentially EC2 cloud instances that run on single-tenant hardware dedicated to a single customer account. They offer certain compliance advantages over standard AWS instances for customers who don’t want to share servers with other customers. But they do not provide the true isolation that customers get on dedicated, bare metal servers. True dedicated servers offer superior performance and customization. And, despite the recent price cuts on EC2 dedicated instances; they still cost more on a total-cost-for-performance basis than do true dedicated servers. Let’s look at the facts:

Servers vs. Instances

It’s important not to confuse AWS EC2 dedicated instances with the true dedicated, bare metal servers offered by Rackspace. Here are some of the key differences:

  1. Isolation: Dedicated servers are completely isolated from the public cloud. You’re on your own hardware with your own network, and you have your own storage. When using EC2 dedicated instances, you are also on your own hardware, but still connected to Amazon’s public cloud. You’re just a dedicated, single-tenant slice of that cloud. If the AWS public cloud suffers an outage, you will be affected. Plus if you need additional block storage, you’re back to multi-tenancy with Amazon’s elastic block storage.
  2. Customization: Dedicated servers are highly customizable. You can choose the operating system, storage, CPU flavor, memory, network speed, etc. In contrast, the specs of an Amazon dedicated instance are predetermined for you, with no option for customization.
  3. Performance: With dedicated servers, your workloads don’t suffer any virtualization penalty. With a dedicated cloud instance, you’re still using a hypervisor, which takes a toll on performance, especially with regard to I/O and CPU. That’s particularly important when running databases. Because you’re still part of the public cloud, dedicated instances must be over-provisioned and over-engineered to design for the higher likelihood of failure due to problems with multi-tenancy, the hypervisor or the host. And while you might eliminate the noisy neighbor of the standard public cloud, you can become your own noisy neighbor as your workloads compete among themselves for the capacity of your dedicated instances.
  4. Reliability: EC2 dedicated instances live on the same class of commodity servers as do regular EC2 instances. Reliability is not improved just because they’re dedicated instances. In some ways reliability may be worse because your dedicated instances will likely be spread across fewer physical servers than if you were using regular instances. Our dedicated servers are backed by network uptime guarantees and industry-leading SLAs and can be customized with the appropriate level of redundancy for the workload.

Unit Cost vs. Total Cost For Performance

To understand the economics of dedicated instances versus those of real dedicated servers, it’s important to consider not only unit cost but the total cost that a customer pays for a required level of performance and support.

Using AWS’ recent price reductions as a guide, let’s do some math. Assume you have one EC2 dedicated instance running for a month. Under its new pricing model, AWS would charge $1,460 a month (based on a 730-hour month), just for the dedicated per region fee — i.e., for the right to spin up and continuously run an instance in one region. Your first Amazon M1.xlarge dedicated instance (15GB RAM), including the per region fee, would run $1,845 a month. On top of that, you would have to pay for bandwidth, support and more.

At Rackspace, we don’t charge a separate fee for each region or for each data center where you have a dedicated server[4]. Nor do we charge extra for our renowned Fanatical Support. Our pricing philosophy is to provide transparency and simplicity. We won’t nickel and dime you with hidden costs, extra charges and complex pricing structures.

Let’s look at the total costs for our dedicated server in a similar configuration: a Rackspace Enhanced 1 Silver dedicated server with 8 CPU cores, 16GB of RAM and two 146GB hard disk drives costs $538 a month (That’s for a managed server with month-to-month pricing, backed by our award-winning Fanatical Support). And our dedicated servers don’t stop there – they offer a level of vertical scaling that isn’t possible with the cloud. For example, you can scale up to 1.5 terabytes (yes, that’s terabytes) of RAM and as many as 32 CPU cores.

A Rackspace customer could get seven dedicated servers (managed with month-to-month pricing) for about $500 a month less than an AWS customers would pay for seven dedicated instances in one region. In that scenario, Rackspace is $3,656 while AWS is $4,158. What’s more, the performance on the dedicated servers would eclipse that of the AWS dedicated instances. And the dedicated servers would be backed, at no extra cost, by Fanatical Support.

One-Size-Fits-All vs. The Best Fit For You

Combining the power of true dedicated hardware with the flexibility and agility of the public cloud is something we’ve done since we launched our pioneering RackConnect[5] hybrid technology in 2009. This combination creates an even more compelling proposition in terms of price for performance than the public cloud alone. Our customer HubSpot, for instance, has stated publicly[6] that when it moved from purely public cloud to the Rackspace Hybrid Cloud, it achieved four times the throughput for the same cost. Not only Hubspot, but also huge organizations like CERN openlab[7] and Fidelity[8] are turning to the hybrid cloud and working with Rackspace to get them there.

By highlighting the importance of dedicated computing, AWS recognized that the hybrid cloud is the future. The Rackspace Hybrid Cloud blends the best of all worlds – dedicated bare metal servers, public cloud and private cloud – to improve performance, price and reliability. And it’s all backed, at no extra cost, by our award-winning Fanatical Support. That’s what our customers consider a real bargain.

Endnotes:
  1. Rackspace Hybrid Cloud: http://www.rackspace.com/cloud/hybrid/
  2. analysts at Wells Fargo and Cowen and Co.: http://blogs.barrons.com/techtraderdaily/2013/07/10/rax-drops-8-fears-of-amazon-pricing-impact-overblown-say-bulls/
  3. every workload belongs in the public cloud: http://www.zdnet.com/amazon-web-services-svp-defends-why-public-clouds-trump-private-models-7000014674/
  4. dedicated server: http://www.rackspace.com/managed-hosting/dedicated-servers/
  5. RackConnect: http://www.rackspace.com/cloud/hybrid/rackconnect/
  6. HubSpot, for instance, has stated publicly: http://www.rackspace.com/blog/rackspace-hubspot-and-fidelity-talk-hybrid-cloud-at-gigaom-structure/
  7. CERN openlab: http://www.rackspace.com/blog/were-collaborating-with-cern-openlab-for-hybrid-cloud-powered-research/
  8. Fidelity: http://gigaom.com/2013/06/19/fidelity-investments-joins-the-openstack-crowd/

Source URL: http://www.rackspace.com/blog/the-value-of-true-dedicated-computing/