Server procurement minimum standards.
Modified on: Tue, 2 Jul 2024 2:35 PMIntroduction
To manage institutional costs incurred from procured equipment, which is difficult or unable to be properly supported, a base set of standards for server and related equipment purchases is outlined here.
These standards are in place to ensure the best value is delivered to the groups procuring such equipment as well as insuring the University avoids liability from incompletely informed purchases.
Applicability
Server equipment to be purchased for use within the University.
Includes Administration, Operations, and Research equipment.
Server Physical Requirements for procurement by the University of Alberta
Exceptions to procurement standards will only be considered for data collection or control computers deployed within research lab environments or infrastructure situations where a computer needs to be in close proximity to operating equipment and these procurement standards are in conflict with needs. Exceptions need to be evaluated on a case-by-case basis. Contact ist@ualberta.ca to start any discussion or exception request.
Servers are to be housed in central University Data Center environments to reduce energy footprint associated with small inefficient hosting spaces scattered around campus and improve the overall quality of the hosting environment. Again, exceptions will only be considered for data collection or control computers associated with research or infrastructure needs.
- Servers are to be procured in rack mount chassis format. Desktop-style chassis are to be avoided or have rack-mount conversion kits with them at minimum if no suitable pure rackmount options are available. Desktop chassis are overall space, energy, and cooling inefficient compared to rackmount designs in general. Physical chassis should be selected for the reasonable lifespan of the equipment and not oversized for unplanned future possibilities.
- Servers should include remote management capabilities (Examples include iDRAC in Dell servers, ILO on IBM and HP servers, Intelligent Platform Management Interface (IPMI), or similar Baseboard Management Controller (BMC) such as OpenBMC). Remote management reduces time to resolve issues, reduces costs of operation and maintenance, and provides the ability to incorporate automation into management activities. Further, hosting locations for servers are based on the most appropriate site based on current site load, server needs, etc. There may be a requirement for travel off North Campus to manage a server if there are no remote management capabilities in the system.
- Servers MUST support power input voltages of 240V. 240V power supplies are more efficient and this lowers the cooling and power costs of server infrastructure. University data center environments will not support 120V-only servers.
-
Servers should have redundant power supplies to ensure continued operation during power outages or maintenance. Single power supplies will be subject to higher levels of disruption as maintenance work is performed on infrastructure.
- Servers should be redundantly network connected to DC infrastructure for service stability and availability during maintenance activities. Servers should leverage Link Aggregation Control Protocol (LACP) to provide this high availability access to network infrastructure or engage IST DNDCS team to discuss alternative options.
- Sizing and design exercises for servers should be undertaken well before procurement. While it is often believed that several smaller servers may be cheaper, in the long run the total cost of ownership can be lower with fewer, larger servers. Design exercises must take application needs along with ongoing operational costs, hardware lifecycle, data lifecycle, etc. into account to deliver the best value for the task to be accomplished. IST teams are available to have these discussions when needed.
- Server connectivity can be accomplished via ethernet interfaces at default speeds of either 1, 10, 25, or 40 Gbps with a typical suggested speed of 10Gbps. 1 and 10 Gbps connections can be supported via Copper while 10,25,40 Gbps options can be provisioned via Fiber Optics. Research applications involving clusters, HPC, or groups of components can use alternatives to Ethernet for interconnectivity inside their systems (i.e. InfiniBand, etc.) but discussions with IST Networking and Data Center staff need to be held prior to procurement to ensure hosting requirements are available and in place.
- High Performance Computing (HPC) equipment of any form should not be procured without sign-off from IST Research Support team and IST Digital Networks and Data Center Services (DNDCS) to ensure that the power, cooling, and network needs for such systems have been appropriately addressed prior to procurement.
- Verify that the computing resources and infrastructure you are acquiring are not from vendors or manufacturers that are prohibited (e.g., Huawei) which can lead to the University being blocked for Canadian and US federal funding opportunities, etc.
Keywords: Server Purchase Buy Procure Standard