[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-08-18。"],[],[],null,["# Best practices for Cloud Storage\n\nThis page contains an index of best practices for Cloud Storage. You can use\nthe information collected here as a quick reference of what to keep in mind when\nbuilding an application that uses Cloud Storage.\n\nIf you are just starting out with Cloud Storage, this page may not be\nthe best place to start, because it does not teach you the basics of how to use\nCloud Storage. If you are a new user, we suggest that you start with\n[Discover object storage with the Google Cloud console](/storage/docs/discover-object-storage-console) or\n[Discover object storage with the gcloud tool](/storage/docs/discover-object-storage-gcloud).\n\nFor best practices for media workloads, see [Best practices for media workloads](/storage/docs/best-practices-media-workload).\n\nNaming\n------\n\nSee [Bucket naming](/storage/docs/buckets#naming) and [Object naming](/storage/docs/objects#naming) for name requirements\nand considerations.\n\nTraffic\n-------\n\n- Perform a back-of-the-envelope estimation of the amount of traffic that will\n be sent to Cloud Storage. Specifically, think about:\n\n - Operations per second. How many operations per second do you expect, for\n both buckets and objects, and for create, update, and delete operations.\n\n - Bandwidth. How much data will be sent, over what time frame? Consider\n using a tool like [Wolfram Alpha](https://www.wolframalpha.com/input?i=350GB+in+5+minutes)\n to avoid mistakes in your calculations.\n\n - Cache control. Specifying the [`Cache-Control` metadata](/storage/docs/metadata#cache-control) on publicly\n accessible objects will benefit read latency on hot or frequently accessed\n objects.\n See [Viewing and Editing Metadata](/storage/docs/viewing-editing-metadata#edit) for instructions for setting object\n metadata, such as `Cache-Control`.\n\n- Design your application to minimize spikes in traffic. If there are clients of\n your application doing updates, spread them out throughout the day.\n\n- When designing applications for high request rates, be aware of\n [rate limits](/storage/quotas) for certain operations. Know the [bandwidth limits](/storage/quotas#bandwidth) for\n certain types of egress and follow the\n [Request Rate and Access Distribution Guidelines](/storage/docs/request-rate). Be especially aware of\n autoscaling and the need to gradually ramp up request rates for the best\n performance.\n\n- When handling errors:\n\n - Make sure your application uses a [retry strategy](/storage/docs/retry-strategy) in order to\n avoid problems due to large traffic bursts.\n\n - Retry using a new connection and possibly re-resolve the domain name. This\n helps avoid \"server stickiness\", where a retry attempts to go through the\n same path and hits the same unhealthy component that the initial request\n hit.\n\n- If your application is latency sensitive, use hedged requests. Hedged requests\n allow you to retry faster and cut down on tail latency. They do this while not\n reducing your request deadline, which could cause requests to time out\n prematurely. For more information, see\n [The Tail at Scale](https://research.google/pubs/pub40801/).\n\n- Understand the performance level customers expect from your application. This\n information helps you choose a storage option and region when creating new\n buckets. For example, consider colocating your compute resources with your\n Cloud Storage buckets for analytics applications.\n\nLocations and data storage options\n----------------------------------\n\nSee the [Storage class](/storage/docs/storage-classes) and [Bucket location](/storage/docs/locations) topics for guidance on how\nto best store your data.\n\nACLs and access control\n-----------------------\n\n- Cloud Storage requests refer to buckets and objects by their names. As a\n result, even though ACLs prevent unauthorized third parties from operating on\n buckets or objects, a third party can attempt requests with bucket or object\n names and determine their existence by observing the error responses. It can\n then be possible for information in bucket or object names to be leaked. If you\n are concerned about the privacy of your bucket or object names, you should take\n appropriate precautions, such as:\n\n - **Choosing bucket and object names that are difficult to guess.** For\n example, a bucket named `mybucket-gtbytul3` is random enough that\n unauthorized third parties cannot feasibly guess it or enumerate other\n bucket names from it.\n\n - **Avoiding use of sensitive information as part of bucket or object\n names.** For example, instead of naming your bucket\n `mysecretproject-prodbucket`, name it `somemeaninglesscodename-prod`. In\n some applications, you may want to keep sensitive metadata in\n [custom Cloud Storage headers](/storage/docs/metadata#custom-metadata) such as `x-goog-meta`, rather than encoding\n the metadata in object names.\n\n- Using groups is preferable to explicitly listing large numbers of users. Not\n only does it scale better, it also provides a very efficient way to update\n the access control for a large number of objects all at once. Lastly, it's\n cheaper as you don't need to make a request per-object to change the ACLs.\n\n- Review and follow [access control best practices](/storage/docs/access-control/best-practices-access-control).\n\n- The Cloud Storage access control system includes the ability to\n specify that objects are publicly readable. Make sure you intend for any\n objects you write with this permission to be public. Once \"published\", data on\n the Internet can be copied to many places, so it's effectively impossible to\n regain read control over an object written with this permission.\n\n- The Cloud Storage access control system includes the ability to\n specify that buckets are publicly writable. While configuring a bucket this\n way can be convenient for various purposes, we recommend against using this\n permission - it can be abused for distributing illegal content, viruses, and\n other malware, and the bucket owner is legally and financially responsible for\n the content stored in their buckets.\n\n If you need to make content available securely to users who don't have user\n accounts, we recommend you use [signed URLs](/storage/docs/access-control/signed-urls). For example, with signed URLs\n you can provide a link to an object, and your application's customers don't\n need to authenticate with Cloud Storage to access the object. When you\n create a signed URL you control the type (read, write, delete) and duration of\n access.\n\nData uploads\n------------\n\n- If you use XMLHttpRequest (XHR) callbacks to get progress updates, do not\n close and re-open the connection if you detect that progress has stalled.\n Doing so creates a bad positive feedback loop during times of network\n congestion. When the network is congested, XHR callbacks can get backlogged\n behind the acknowledgement (ACK/NACK) activity from the upload stream, and\n closing and reopening the connection when this happens uses more network\n capacity at exactly the time when you can least afford it.\n\n- For upload traffic, we recommend setting reasonably long timeouts. For a good\n end-user experience, you can set a client-side timer that updates the client\n status window with a message (e.g., \"network congestion\") when your\n application hasn't received an XHR callback for a long time. Don't just\n close the connection and try again when this happens.\n\n- An easy and convenient way to reduce the bandwidth needed for each request is\n to enable gzip compression. Although this requires additional CPU time to\n uncompress the results, the trade-off with network costs usually makes it\n very worthwhile.\n\n An object that was uploaded in gzip format can generally be served in gzip\n format as well. However, avoid uploading content that has both\n `content-encoding: gzip` and a `content-type` that is compressed, as this\n may lead to [unexpected behavior](/storage/docs/transcoding#gzip-gzip).\n- We recommend using [resumable uploads](/storage/docs/resumable-uploads), which allow you to resume\n transferring data even when a communication failure has interrupted the flow\n of data. You can also use XML API multipart uploads to upload parts of a file\n in parallel, which potentially reduces the time to complete the overall\n upload.\n\nDeletion of data\n----------------\n\nSee [Delete objects](/storage/docs/deleting-objects) for guidelines and considerations on deleting data.\nYou can also use [features for controlling data lifecycles](/storage/docs/control-data-lifecycles) to help protect\nyour data from getting erroneously deleted by your application software or\nusers."]]