미래용 예약을 사용하면 설정된 기간 동안 용량을 예약하도록 요청할 수 있습니다.
예약 요청을 만든 후에는 검토를 위해 Google Cloud 에 제출해야 합니다. Google Cloud 에서 요청을 검토하는 데는 일반적으로 5일이 걸립니다. 요청이 승인되면 Compute Engine은 선택된 날짜와 시간에 요청된 용량으로 온디맨드 예약을 만듭니다. 그러면 사용자는 이러한 예약을 사용하는 컴퓨팅 인스턴스를 만들어 예약을 사용할 수 있습니다.
예약 기간이 종료된 후에는 예약을 수정하거나 삭제할 수 있습니다.
캘린더 모드에서 미래용 예약을 사용하여 미리 정의된 기간 동안 TPU를 예약할 수 있습니다. 이 유형의 예약을 만들려면 먼저 선택한 수와 유형의 TPU를 리전에서 사용할 수 있는 시점을 확인합니다. 그런 다음 사용 가능한 것으로 확인된 속성으로 예약 요청을 만들고 검토를 위해 요청을 제출합니다. Google Cloud 에서 1분 이내에 요청을 검토하고 요청을 승인하면 Compute Engine에서 다음을 수행합니다.
Compute Engine에서 온디맨드 예약을 만듭니다.
Compute Engine에서 네트워크 지연 시간을 최소화하기 위해 요청된 리소스를 최대한 서로 가까운 위치에 예약합니다.
예약 기간이 시작되면 TPU 인스턴스를 만들어 예약을 사용할 수 있습니다. Compute Engine은 지정된 예약 기간이 끝나면 예약과 예약을 사용하는 인스턴스를 삭제합니다.
[[["이해하기 쉬움","easyToUnderstand","thumb-up"],["문제가 해결됨","solvedMyProblem","thumb-up"],["기타","otherUp","thumb-up"]],[["이해하기 어려움","hardToUnderstand","thumb-down"],["잘못된 정보 또는 샘플 코드","incorrectInformationOrSampleCode","thumb-down"],["필요한 정보/샘플이 없음","missingTheInformationSamplesINeed","thumb-down"],["번역 문제","translationIssue","thumb-down"],["기타","otherDown","thumb-down"]],["최종 업데이트: 2025-07-30(UTC)"],[],[],null,["*** ** * ** ***\n\nThis document describes the different reservation types that you can use to\nreserve capacity for Compute Engine instances. To learn more about the\nresources that you use to create instances, see\n[Compute Engine instances](/compute/docs/instances).\n\nReservations help ensure that you have the available resources to create\ninstances with the same hardware (memory and vCPUs) and optional resources\n(GPUs, TPUs, or Local SSD disks) whenever you need them. Reservations offer the\nfollowing benefits:\n\n- **High assurance of capacity**: you reserve resources to accommodate for\n future increases in demand, such as the following:\n\n - Growth\n\n - Planned or unplanned spikes in usage\n\n - Large migrations\n\n - Backup and disaster recovery\n\n- **Exclusive access**: reservations prevent others from using your reserved\n resources.\n\n- **Inherited properties**: reservations inherit the same properties as your\n chosen machine family.\n\nAfter you reserve capacity, you can use it to create instances that match the\nreservation. You don't incur any additional charges when you create these\ninstances. You only pay for resources that aren't part of the reservation, such\nas disks or IP addresses.\n\nLimitations\n\nAll reservation types have the following limitations:\n\n- Reservations are\n [zone-specific resources](/compute/docs/regions-zones/global-regional-zonal-resources#zoneresource).\n\n- You can't use your reserved capacity to create the following\n Compute Engine resources:\n\n - Spot VMs or preemptible instances\n\n - Sole-tenant nodes\n\nChoose a reservation type\n\nThe following diagram helps you choose the Compute Engine reservation\ntype that best fits your workload's needs:\n\nThe questions in the preceding diagram are as follows:\n\n1. Do you need capacity right away?\n\n - **Yes**: Go to the next question.\n\n - **No**: Go to question 3.\n\n2. Do you need flexibility on how long to hold capacity?\n\n - **Yes** : See [Use on-demand reservations](#on-demand-reservations).\n\n - **No**: Go to the next question.\n\n3. Do you need high-demand resources like GPUs?\n\n - **Yes**: Go to the next question.\n\n - **No** : See\n [Use future reservations](#future-reservations).\n\n4. Do you need resources for more than 90 days?\n\n - **Yes** : See\n [Use future reservations in AI Hypercomputer](#reserve-capacity-aihc).\n\n - **No** : See\n [Use future reservations in calendar mode](#future-reservations-calendar-mode).\n\nUse on-demand reservations\n\nWith on-demand reservations, you can reserve capacity for compute instances and\nget it as soon as you reserve it. After you create an on-demand reservation, you\ncan consume, modify, or delete it whenever you need to.\n\nFor more information, see\n[About reservations](/compute/docs/instances/reservations-overview).\n\nUse future reservations\n| **Preview**\n|\n|\n| This product or feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA products and features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nTo reserve instances for a set period, you can use future reservations. After\nyou create a reservation request, you must submit it to Google Cloud for review.\nGoogle Cloud typically takes five days to review your request. If your\nrequest is approved, then Compute Engine creates on-demand reservations\nwith your requested capacity on your chosen date and time. To consume these\nreservations, you create compute instances that use the reservations. After the\nreservation period ends, you can modify or delete the reservations.\n\nFor more information, see\n[About future reservation requests](/compute/docs/instances/future-reservations-overview).\n\nUse future reservations in calendar mode\n| **Preview**\n|\n|\n| This product or feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA products and features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nTo reserve GPU instances or TPUs for up to 90 days, you can use future\nreservations in calendar mode. To create this type of reservation, first view\nwhen your chosen number and type of resources are available in a region. Then,\ncreate and submit a reservation request with the properties that you confirmed\nas available. If you can successfully create the request, then\nGoogle Cloud approves it within a minute. After the request is approved,\nCompute Engine does the following:\n\n- Compute Engine creates an on-demand reservation.\n\n- Compute Engine reserves your requested resources as close to each\n other as possible to minimize network latency.\n\nAt the start of your reservation period, you can consume the reservation by\ncreating GPU or TPU instances. At the end of the reservation period,\nCompute Engine deletes the reservation, and stops or deletes any\ninstances that consume the reservation based on the termination action that you\nspecified for the instances.\n\nFor more information, see\n[About future reservation requests in calendar mode](/compute/docs/instances/future-reservations-calendar-mode-overview).\n\nUse future reservations in AI Hypercomputer\n| **Preview**\n|\n|\n| This product or feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA products and features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nContact your account team and request to reserve GPU instances for large-scale\nartificial intelligence (AI) and machine learning (ML) workloads. After Google\ncreates a draft reservation request for you, submit it for review if everything\nlooks correct. Google Cloud immediately approves the request, and then\nCompute Engine does the following:\n\n- Compute Engine creates on-demand reservations.\n\n- Compute Engine reserves your requested resources as close to each\n other as possible to minimize network latency.\n\n- Compute Engine reserves resources with topology-aware scheduling, as\n well as enhanced monitoring and maintenance.\n\nAt the start of your reservation period, you can consume the reservation by\ncreating GPU instances. At the end of the reservation period,\nCompute Engine deletes the reservation, and stops or deletes any\ninstances that consume the reservation based on the termination action that you\nspecified for the instances.\n\nFor more information, see\n[Reserve capacity](/ai-hypercomputer/docs/reserve-capacity) in the\nAI Hypercomputer documentation."]]
