IT managers looking for dollars to fund new projects may be able to find them by critically evaluating their existing on-premises systems. Research by TSO Logic, based on analysis of 10,000 customer physical servers and 25,000 virtual machines, suggests that 45% of existing VMs would run more efficiently in the cloud.
A secondary finding was that merely upgrading systems to the latest generation of servers, whether on premises or in the cloud, would also result in considerable savings. Twenty-five percent of the physical servers surveyed were more than three years old, meaning they dated back into the Intel Nehalem chip era of 2013 and its predecessors.
Nehalems were typically built with 45 or 32 nanometer circuits, tiny for their time but currently outstripped by the 14 nanometer circuits of the Broadwell chip set that were mass produced in 2015 and 2016. Both chips in their respective times served as the basis for the latest Xeon server lines. Each progressive reduction in circuit size leads to higher transistor counts on the chip, faster operations and reduced electricity consumption to accomplish the same amount of work.
In data center servers, Nehalems are a fifth generation chip; Broadwells are ninth generation chips, with some Broadwells appearing in the Xeon server line, some in laptops and some in high performance servers. In between were sixth-generation Sandy Bridge, seventh-generation Ivy Bridge, and eighth-generation Haswell CPUs.
Given the higher speeds and greater processing power of the ninth generation chips, analytics firm TSO Logic concluded: "The same workloads running on Generation-5 servers could run on 30% fewer Generation-9 servers." By the same workloads, TSO Logic means the same resources provided to each operating system deployed or a "like for like" exchange, it said in a recent announcement.
Cutting the number of servers leads to savings on several fronts, including reduction in the number of operating system and other software licenses and reduced consumption of electricity, as well as lower management costs.
But reducing the server footprint by 30% isn't the only area in which IT managers can mine savings in their existing data centers, the research concluded, according to TSO Logic's CEO Aaron Rallo in an interview.
Is there really a need for migration assessment? Gartner Sees $1 Trillion Spending Shift To The Cloud.
Twenty-six percent of the virtual machines it inspected on-premises were over-provisioned by their creators, Rallo said. Migrating them to the cloud and right-sizing them there would result in a 36% savings over their current operating cost, he said.
When both things are done at the same time – modern hardware replaces older models and workloads get right-sized – then Broadwell-based servers can replace a set of servers that are three years old or older with 54% fewer servers that will be able to do the same amount of work, according to the Dec. 15 announcement of the research findings.
Rallo said cloud providers tend to refresh their hardware to latest generations more frequently than some enterprise operations included in TSO Logic's survey. Workloads sent to the cloud can be right-sized through tools typically available there. In addition, right sizing can be done on premises with third party systems, including TSO Logic's Match and TSO Metrics. TSO Logic became a partner Nov. 22 with Amazon Web Services to provide its tools to help potential customers plan their migrations to the cloud. For five years, TSO Logic has been a partner with Intel in helping customers assess when they should upgrade servers. Its product TSO Match can assess what's running where and what amount of modern server power is needed to replace it.
When TSO said 45% of VMs would run more efficiently in the cloud, it didn't specify which cloud. Although there are pricing variations between them, Rallo said all estimated expenses are based on an average of the two top suppliers, Amazon Web Services and Microsoft Azure. Their prices remain similarly high, he said.
Rallo said TSO Logics' analysis of the research data wasn't meant to serve as a precise budgeting mechanism but could be used in decision-making on what savings materialize in either upgrading with new hardware or migrating to the cloud.
Use of Amazon T2 series instances, a small memory and CPU virtual server that can be increased in size for "bursts" of processing power, figured prominently in the estimates, Rallo said. The T instances were launched by AWS on July 1, 2014, and are its smallest, general purpose offering. The largest T2, a double extra-large comes with 32 GiB of memory, the maximum, versus .5 GiB of memory for the introductory, micro T2. The GiB or Gibibyte is closely related to Gigabytes but is larger. One Gibibyte equals 1.074 GB.
TSO's use of the T's would tend to maximize the expense savings, but not necessarily match all the workloads that were covered by the TSO Logic survey. A database application, for example, would be likely to require greater memory than what's available with the T instances and might be launched using a larger general purpose instances, such as the M3 or M4 instances, observers not connected with the research might say.
Nevertheless, few third parties have as much experience as TSO Logic in delivering analysis of how servers can be exchanged for something else. The Vancouver company was founded in 2010 to provide a platform of applications that could optimize IT operations. It received the Uptime Symposium's Green Enterprise IT Award in 2014.
The TSO Logic Platform collects data from multiple data center systems and can use it for the analysis of what's running where and what's the right size for a given application. Rallo said the his firm used data from the Nagios open source system management system, VMware's vSphere and vCenter, Datadog monitoring and metric collection, Solar Winds network and system management, and other systems.
"All the data we needed already existed," Rallo said in the interview. "All that was needed was the analysis." He added that TSO Logic's analysis engine "is not pre-disposed that everything goes into the cloud," he said.Charles Babcock is an editor-at-large for InformationWeek and author of Management Strategies for the Cloud Revolution, a McGraw-Hill book. He is the former editor-in-chief of Digital News, former software editor of Computerworld and former technology editor of Interactive ... View Full Bio