18 December 2013 Posted by Andrew Bohling
Many public clouds have been built on commodity hardware with the assumption that, by virtualizing the infrastructure, it doesn’t matter what kind of hardware is used. Cloud-based compute and storage services have given organizations the false impression that better-performing, more scalable instances don’t rely on differences in underlying hardware. This is one of the greatest misunderstandings of infrastructure as a service (IaaS), the idea that hardware innovations don’t bolster the performance or cost-effectiveness of a chosen solution – or that application developers don’t care about what hardware is behind the service. The truth is that hardware isn’t just a commodity.
With most public cloud services, software defines much of the value that customers receive. Automation, control, visibility, and management features help cloud users do more with the resources at hand. A drawback of cloud services built on the most basic, vanilla hardware is that they focus too much on cost optimization and not enough on application-specific requirements. Cloud customers, particularly developers, may need solutions that cater to unique needs and provide more specific functionality and optimization than what less capable hardware can offer.
This November, Amazon Web Services (AWS) announced the availability of new C3 instances, a line of compute-optimized EC2 instances that use SSD-based storage and the latest generation of Intel processors. They offer the best price to performance ratio of any current EC2 offering. C3 instances are engineered to improve inter-instance latency, lower network jitter, and provide higher packet per second performance – and specialized hardware using SSDs plays a key role in making this functionality possible.
The industry is seeing great demand for these high powered, custom hardware solutions, so much that AWS is experiencing a shortage of C3 instances. Their chief evangelist, Jeff Barr, said that in only two weeks C3 usage had exceeded the level that the former fastest-growing instance had achieved in twenty-two weeks time. With magnitudes better performance and better price/performance ratios, it proves that hardware still matters.
There are other areas where demand for more capable or specialized hardware solutions is increasing as well. For example, organizations that use large relational databases need fast hardware to achieve massive scale and performance. These databases must “scale up” on a single high-performance server rather than “scale out” over multiple generic servers. For cloud deployment, they require the largest instance size possible, which in turn requires the most powerful underlying hardware.
Hardware is also still the key to maximizing network performance. For example, 1 Gbps networking is typically not enough to meet the performance requirements of low-latency applications running in shared multi-tenant environments. Network hardware that supports 10Gbps networking is often needed to ensure that tenants don’t impact each other’s performance, particularly when a high-percentage of I/O travels over the shared LAN to an SAN.
Some cloud computing benefits, such as massive scale and low costs, are made possible by using inexpensive, generic hardware. But cloud users are starting to realize that services built on the most basic hardware don’t meet their needs for every application. More capable, higher-performance hardware is often needed to ensure applications run properly, whether in a public cloud environment or a traditional data center. To get what your applications need from a public cloud, it pays to look at what’s under the hood.
]]>
Recent blog posts
- Revisiting the Benefits of SaaS
- Compliance Drives Buying Decisions in the Cloud
- Changing Competitive Dynamics for IaaS Providers
- Good News for Cloud Job Seekers in 2015
- Colocation: The New Center of Hybrid IT
- The Problem with Forced Cloud Reboots
- OnApp Federation Grows Stronger with SolusVM Acquisition
- HP to Acquire Eucalyptus: Questions, Comments and Concerns
- Is It Really ‘Game Over’ for Most IaaS Providers?
- The Big Cloud Winners Are…Not Just the Hyperscale Players