Open commodity servers have reached the point where the cost of server deployment, management, facilities, and recycling exceed the acquisition costs. Two major trends have occurred to address this trend:
- Converged infrastructure - a complete package of servers, networking and storage delivered as a working system as a single SKU;
- Managed infrastructure - a set of infrastructure that provides all the data and software to be managed with minimal effort.
The HP ProLiant Gen8 server line is managed infrastructure, and is a major upgrade in managing racks of servers and disks with greater efficiency, ease-of-operation, and ease-of-asset-management. HP invested in an intense two-year program to identify and reduce operational human error and automate common processes in the deployment, servicing, operation, and recycling of commodity servers. At the same time HP has focused on reducing the environmental impact of servers and reducing the facilities cost of housing server racks. The result has been a large number of small changes, which collectively add up to significant savings in the server and overall IT budgets.
Wikibon utilized its standard business and IT models to assess the overall impact. Wikibon found that compared with the HP ProLiant Gen7 servers, the impact on the server budget would be a yearly reduction of about 15%, and the impact on the overall IT budget would be about 4%. For a company of 4,000 people with a revenue of $1 billion, and an IT budget of $30 million (3% of revenue), the saving would be about $1.2 million. A three-year business case would have net present savings of $3.3 million.
The ability to realize these savings will be a function of the control that IT has over all the IT budget line items. Organizations where IT does not have responsibility for facilities costs or has outsourced servicing and management of server infrastructure will see far lower direct savings but should push for lower prices from their vendors.
Impact on IT Budgets
Wikibon uses a standard IT cost model for financial analysis. Wikibon models a medium-sized business, with:
- $1 billion in revenues,
- 4,000 employees,
- $250,000 revenue per employee,
- IT budget of 3% of revenue.
The budget line items are shown in the first column of Figure 1, the percentage of the total IT budget is shown in the second column for each line item, and the cost in the third column. The fourth column is the cost of the line item assuming HP ProLiant Gen 8 servers, HP ProLiant Gen8 racks and full implementation of software and processes to take advantage of the new platform. The fifth column is the percentage saving against the total IT budget line, and the last column are notes on the types of savings included.
The savings in Figure 1 come from the following areas:
- Facilities - The facilities cost savings are based on a large number of small items including:
- Much improved power supplies (94% efficiency, with better power curve characteristics and improved harmonic distortion),
- Improved fan optimization with more sensors on the all the components and racks,
- Improved airflow through the rack mesh to 81%, reducing the fan power required,
- Reduced voltage required for memory DIMMs from 1.5v to 1.35v,
- Improved baffle design, cable clipping, disk design to improve airflow,
- Improved rack design, with improved PDUs, cable management, sizing to fit ProLiant Gen8 servers,
- Feedback on processor and disk utilization and understanding of all the heat sources, enabling more accurate setting of fans, and ability to predict and avoid hotspots,
- Feedback on utilization, allowing true knowledge of power requirements and optimization of PDUs and disaster cutover (leads to lower power provisioning costs),
- Discovery services that understand the age and power profile of server resources. The knowledge of assets and utilization leads to the ability to optimize and power down underutilized resources.
One of the areas not addressed was any increase in operating temperature range, which would have reduced cooling costs and allowed greater use of ambient air. The upper range is still 35C.
- Server Infrastructure - the direct savings in server infrastructure come from:
- Better tracking of assets - uses a "One Button" strip on the left side of the rack, which can read the server ID on the side of the server. This discovery service allows automatic asset management of the servers and access to the configuration, age, warranty, and utilization information from the data center software. Improved tracking leads to better decisions on deployment of servers against workloads, fewer servers and lower server acquisition costs.
- Operations - the operational cost savings come from a reducing the effort required in install, deploy, maintain, and recycle rack mounted servers. Some of the new design features include:
- Reduction in time to deploy servers into racks from ten minutes to five minutes from no tools required, easier unpacking, standard rails, ease of attaching cables, and much improved labeling.
- Smart checking of installation against best practice guideline - for example, connecting of power cords to different PDUs to ensure availability.
- Significant reduction in the time-to-service servers, including a smart socket guide to install new processors with no risk of bending a pin (HP is taking the gambling highs from service engineers), LED lights on disks in service with a "do not remove" sign write large, reducing the risk of the wrong disk being pulled (the highest cause of unplanned outages), improved cabling, and improved signs.
- Perhaps the most important aspect of improved maintenance is the closed loop system among the development team, the manufacturing team, the huge amount of data from the servers and disks, and the service system that analyses and feeds the data back to HP service engineers, partners who want to provide their own services, and back to the development team. All of this data can be focused on a continuous learning capability for better design, better components, and better processes throughout the lifecycle of the servers.
- The reduction in the number of servers has an impact on the operational software costs (very little impact on development or package software).
Figure 2 shows just the server infrastructure components of the IT budget, which totals about a quarter of the IT budget. The same savings as in Figure 1 are applied to this subset of the IT budget, with a saving of approximately 15% of the server budget.
The business case presented in this alert is preliminary, and Wikibon members are encouraged to add their comments and suggestions to the types and size of benefits that ProLiant Gen8 will bring to different classes of user. There are some areas of weakness - Wikibon would have hoped for increased operating temperatures, for example. Some of the claims have been over-stated: for example, the saving assumed that the amount of server capacity would go down year-on-year. As Moore's law is applied, the elastic effect of lower prices per unit of computing are translated into high function from increased software and increased system deployment. Some areas have been understated: for example, the challenge of putting in high-availability processes and procedures without the features that ProLiant Gen8 brings to the table means that the operation costs should be lower.
Another area of overstatement is the degree of confidence that patterns can be identified on how applications are working and applied to optimize performance. There is neither the data nor the architecture to provide true application SLA monitoring. Wikibon believes this should be a strong focus for ProLiant Gen9. Server administration will need to migrate to management of virtual machines and applications.
Action Item: HP has invested heavily in the ProLiant Gen8 industry standard server platform, and has significantly improved the OPEX components of managing large numbers of servers, both rack and blade. HP has reduced the management difference between networked servers, such as EMC's VCE and Cisco's UCS offerings, while simultaneously providing a much richer number of server options and functionality. HP server and infrastructure customers in particular will be well pleased with the announcement. CXO's should understand that in order to optimize the value of ProLiant Gen8 server infrastructure, HP server racks as well as control over the facilities budget are very desirable. Investments also need to be made to take advantage of the torrent of data that will emanate from this best-of-breed managed infrastructure.
Footnotes: See more coverage of HP's Project Voyager/Gen8 here.