Skip to content
My.Environment

HYAK Extended Use Agreement

to

Hyak High Performance and Data Ecosystem Memorandum of Understanding

November 30, 2023 – v1.1

  College of the Environment                                                                      

Sponsoring Organization

Purpose of the Node Retirement and Extended Access Addendum

This Addendum to the Hyak HPDE Memorandum of Understanding (MOU) dated January 31, 2023 revises the retirement policy of nodes (AKA “slices”). This revised policy will be effective January 1, 2024.

This revision offers a more flexible retirement policy to allow continued access to Hyak once a node has been in use for four years, with an associated Extended Access Fee (Access Fee) to cover the costs associated with operating the computing equipment to support that Access, as well as the costs of operating older equipment, and which will operate under the terms and conditions outlined in this Addendum.

Rationale for new policy

Hyak operates in a ‘condo’ model, whereby slices are purchased by users and provide the computing equipment in the system, and the slices operate within the Hyak infrastructure. The cost of supporting the infrastructure and the operation of Hyak must be recovered, which is accomplished in part through Sponsorship Commitments and associated Slot Fees, and through Self-Sponsored Slot Fees. The cost of supporting older nodes goes up as time goes on, which includes both hardware repair and labor costs. In addition, the “lost opportunity” cost of using the limited data center resources goes up significantly with older equipment, because less computing can be accomplished with older equipment for the same power consumption as newer equipment. The increasing lost opportunity cost is reflected in an annual increase in Extended Access Fees.

Terms for Revised Node Retirement Policy 2.1

1. Node Ownership and Extended Access. We will continue to adhere to the original terms for ownership of Hyak nodes as described in Sponsorship MOUs: after four years, Node Owners will abandon their node(s) in place, and ownership reverts to the Hyak System. At that time, Node Owners will be eligible for ongoing access to Hyak, as a Participant, beyond that time for a fee, and as long as resources are available in the Hyak System to support such “Extended Access”. This Extended Access (Access) includes on-demand access to a node of equivalent or better computing power to the previously owned node, and equivalent priority for the use of the Hyak Checkpoint Queue. This Access will be offered for a maximum of four years following retirement of the node (a total of eight years after deployment), and may terminate sooner, as described in the Termination clause. An Extended Access Agreement will be executed between the Hyak Operator, the Participant, and with their Sponsor, if any. Access rates are relatively small for the first two years of Extended Access, and rise significantly in the following two years.

  • Cost. The Fee for Extended Access will cover the costs associated with providing Extended Access, as well as proportionate support of the entire Hyak High Performance and Data Ecosystem (Slot Fee). The Fee will increase annually, starting relatively small for the first two years, and significantly higher for the following two years. This increase reflects the lost opportunity costs and the increased operational cost of using the Hyak Ecosystem to support aging equipment. Details of the costs based on the Sponsor’s local policy are described in the Sponsor’s Cost Policy section, below.
  • Termination. Participants may terminate an Extended Access Agreement at any time, which will be effective the first day of the month following notice to UW-IT (help@uw.edu).

A Sponsor may elect to terminate their support discount for any reason, with three months notice to the Participant and UW-IT.

The Hyak Operator (UW-IT) may elect to terminate the Extended Access Agreement for any reason, with or without prior notice. Typically this would only happen due to loss of appropriate node resources in the system (e.g., for hardware failures that cannot be repaired in a cost-effective manner – no prior notice given), to retirement of underlying cluster infrastructure (see clause below), or due to overall system capacity constraints, such as data center capacity (with as much notice as feasible).

After four years, the Extended Access Agreement will be terminated.

  • Infrastructure Retirement. If the cluster infrastructure (e.g., MOX) supporting the Extended Access reaches a point where the Hyak Operator determines that the infrastructure should be retired, we will enter an “Infrastructure Wind-down” period. That determination could be based on, but not limited to the following: hardware or software support is no longer available for a critical component of that infrastructure, service or repair costs would be unacceptably expensive, security breaches are identified that cannot be adequately addressed, available electrical power to support Hyak is overly constrained and does not permit addition of new nodes, or for some other significant or critical system-level reason. The System Operator will make a good faith effort to provide a minimum of three (3) months notice before entering a Wind-down Period. The expected duration of the Wind-down Period will be included in that notice. The entire affected infrastructure will be retired at the end of the Wind-down Period, and any Extended Access supported by that Infrastructure will be terminated at that time. If the supporting infrastructure experiences a hardware failure during the Wind-down Period and repair is impractical or overly expensive, or if a significant security risk is identified that cannot be adequately addressed, or similar operational issue arises, all Extended Access Agreements dependent on that infrastructure will be immediately terminated.
  • Sponsor Support for Extended Access. For Owners of Sponsored Nodes, the Sponsor may elect to sponsor Extended Access for that Owner, or not. Furthermore, if the Sponsor will continue to sponsor use of the system, they may elect to pass on some, all, or none of the Sponsor’s related costs (Sponsor’s Annual Slot Fee) to the Participant, which will be reflected as a discount to the Sponsor’s Slot Fee, and included in the Extended Access Fee. If the Sponsor does not elect to continue sponsorship, Extended Access will be considered to be Self-Sponsored, with a higher cost to the Participant based on the Self-Sponsored Slot Fee. The System Operator will establish an Addendum to the Sponsorship MOU with each Sponsor to document the Sponsor’s local policies for supporting Extended Access, including the amount of the financial support the Sponsor will provide for each year of access. The Operator will transfer to the Sponsor on a monthly basis the monthly amount of their support cost (the Sponsor’s Slot Fee) that the Sponsor has determined will be paid by the Participant.

Sponsor may change the discount percentage with three month’s notice to UW-IT and all sponsored Participants.

For Self-Sponsored Nodes, the System Operator (UW-IT) will serve as the Sponsor-of-record and make the decision whether or not to offer Extended Access to the former Node Owner. If Extended Access is approved, standard Unsponsored Slot Fee rates will apply.

  • Exceptions. Any exceptions to these Terms, or to the specific rates may be discussed with the Administrative Official of the Office of Research Cyberinfrastructure (ORCI), the Associate Vice Provost for Research Cyberinfrastructure.

Sponsor’s Cost Policy

An Extended Access Fee (Access Fee) will be charged directly to the Participant (PI) by the Hyak System Operator (UW-IT) on a monthly basis, which covers the cost of supporting the resources necessary to support the extended access. A Sponsor may agree to continue to support the Participant’s use of the system during Extended Access, or may decline to entirely, and that decision may vary annually during a Participant’s Extended Access. If the Sponsor will continue to support the Participant’s use of Hyak during Extended Access, the computing equipment (slice or slices)

required to support Extended Access will be treated as operating within the Sponsor’s Slot Allocation. Furthermore, the Sponsor may elect to share none, some, or all of the sponsorship costs associated with the slot(s) used to support the slice(s) – their Annual Slot Fee amount – and those support costs may vary annually. The support costs not covered by the Sponsor (the

“Non-discounted Support Cost”) will be the responsibility of the Participant, and will be included in the Access Fee. If the Sponsor declines to support the Participant’s Extended Access, Extended Access will be charged at the Self-Sponsored rates.

The rate schedule shown in Table 1 indicates whether the Sponsor will sponsor Extended Access for the Participant, on a year-by-year basis, in the column labeled “Sponsored?”. The Table also indicates the percentage of the associated Slot Fees that the Sponsor will cover; the balance of the Sponsor’s Slot Fee (the Non-discounted Support Cost) will be included in the Participant’s Access Fee, and then subsequently transferred to the Sponsor’s financial account, as identified in Table 1.

This rate schedule will be included in the initial notice given to the Owner of an impending Node retirement, and in their subsequent Extended Access Agreement.

Rate Adjustments. This rate schedule for Extended Access can be changed, with three months notice to Participants and Slice Owners.

Grace Period during Initial Year

In order to transition to this new policy, a grace period will be granted to certain Participants. Extended Access Fees associated with any Node that continues to be sponsored under the CoENV agreement will be waived until December 31, 2024 or the end of seventh year of service for that Node, whichever comes first. In other words, any nodes that were deployed on or after January 1, 2018 will be exempt from all Extended Access fees through December 31, 2024. And any nodes that were deployed during 2017 will owe no fees for a commensurate portion of 2024 that ends when

the node reaches seven years of service. Beginning January 1, 2025, Extended Access Fees will be charged to all Participants, based on the Rate Schedule in Table 1.

Process

The Hyak System Operator (UW-IT) will contact individual Node Owners and their Sponsor at least two months before their Node(s) will reach the end of their Standard Service Life (four years), to notify them that retirement is near. Information will be provided to describe the terms and the costs for Extended Access, based on the Sponsor’s local policies and the other terms in this Addendum.

If the Node Owner requests Extended Access under those terms and costs, an Extended Access Agreement will be executed between the Node Owner, Sponsor (if any), and UW-IT. If the Node Owner declines or does not request Extended Access, the Node Owner will not have access to Hyak based on that Node once it reaches four years of service, and the Node Owner will have two additional months to extract any data associated with that node. An Extended Access Agreement may encompass more than one node, if they are all to be treated identically.

Once the node does reach its four-year life, title will be officially transferred to UW-IT through the UW Equipment Inventory System, with notification to the original Node Owner and to the Sponsor. If the Owner has requested Extended Access, the budget number (“Cost Center” or “Worktag”) supplied by the Participant in the Extended Access Agreement will be charged on a monthly basis, with the Non-discounted Support Costs transferred to the Sponsor’s financial account as identified in Table 1.