GPU 加速分析速度快，但全天候运行成本高昂。本文介绍如何利用云数据仓库（cloud data warehouse）中的空间分析功能，以更低的成本处理数十亿个空间点。

GPU accelerated analytics is fast but costly to run 24/7. See how spatial analysis in your cloud data warehouse can handle billions of points for far less.

对大量地理空间数据进行分析与可视化，对许多从事GIS或地理空间分析的分析师而言，有时近乎一种遥不可及的梦想。文件大小限制，以及难以支撑大规模空间分析的基础设施，始终让这一目标显得触手难及。本文将重点探讨两种应对路径：GPU加速分析与具备空间能力的云数据仓库。 若您曾见过由GPU加速数据库或渲染引擎驱动的地图或仪表板，便会发现这类仪表板响应迅速，可流畅呈现海量数据。它们能够轻松存储、查询并可视化数百万乃至数十亿行数据，且无需对数据进行大量预处理或转换。尽管GPU加速分析在分析与可视化速度上优势显著，但此类服务亦存在若干缺陷，其中最主要的是高昂成本与持续运行开销。 在深入讨论具体细节之前，我们先用一个类比帮助理解：假设您正考虑是否购买一辆汽车。您居住在城市中，日常通勤主要依赖共享单车、公共交通以及Uber等网约车服务。借助这些替代方案，您可在需要时即时获取恰如其分的服务，并仅按实际使用量付费；若需更优质服务（例如升级车型或更快接驾），也可选择承担相应附加费用。GPU好比自购汽车，而云数据仓库则类似使用公共交通/单车/网约车。 当您采用GPU加速地理空间服务时，GPU必须全天候持续运行——无论您是否正在使用。随着数据规模扩大或新增仪表板，为维持性能，您可能还需增加GPU数量，而此类服务价格不菲。单看云基础设施成本即可见一斑：以AWS EC2 p4d.24xlarge实例（AWS所提供最强大也最昂贵的实例之一）运行GPU加速服务，其标价约为每小时42.93美元。该估算尚未计入软件许可费及其他潜在成本。 您可手动启停服务以实现按需伸缩，或配置自动扩缩容机制。但额外增加的运维层本身亦需投入时间与成本。此外，若在同一服务器上部署其他仪表板或服务，因并发查询增多可能导致性能下降，您可能还需另行配置更多服务资源。若您已在使用数据仓库并将数据同步至GPU加速平台，则不仅需支付数据仓库中的数据存储与查询费用，还需额外承担GPU服务费用。而对于仅存于GPU数据库中的数据，您将丧失将其与数据仓库中其他数据集成的能力。 数据仓库正日益普及并持续提升可扩展性，已成为管理与分析海量地理空间数据的理想方案。它提供低成本、高效率的存储，以及免运维的无服务器（serverless）或弹性计算资源。这意味着您仅需在执行查询时调用所需算力，且仅按使用时长付费。 如今，这一原生范式已进一步演进为“智能体GIS”（Agentic GIS）。CARTO AI Agents能够理解自然语言，并直接在您的数据仓库中执行空间任务——分析师只需提出问题，即可获得地图或答案，无需手动编写每一条查询语句。关于为何数据仓库是空间分析的理想载体，详见我们此前发布的文章《为何在地理空间分析中采用数据仓库》。 目前大多数GPU加速工具均以栅格格式（如PNG瓦片）生成地图瓦片。相较之下，矢量瓦片代表一种更现代的方法，支持借助三维渲染、弧线、动画等实现更高级的可视化与分析能力。CARTO云原生平台在地图瓦片生成与可视化方面，较GPU加速分析工具具备若干关键差异化优势。 采用此方法的主要权衡在于：您将失去HEAVY AI仪表板所具备的动态逐点过滤与实时重绘能力。不过，借助精巧的SQL语句与选择性计数，您仍可在聚合视图中实现部分此类过滤功能。首张地图实际使用了两个瓦片集：在缩放级别17以上显示基于点聚合的视图；在17以下则展示原始数据。您可在Builder中设置基于缩放级别的过滤逻辑，并据此配置瓦片集以控制不同缩放层级下的呈现效果。 H3地图则利用一项新功能生成空间索引瓦片集。一旦为数据分配H3或Quadbin空间索引，即可仅凭该索引值生成瓦片，无需几何对象（geometry），因而远快于其他瓦片生成方式。 真正的成本节约体现在此处：若每月仅执行一次上述查询，并将相应数据与瓦片持久化存储，本用例的总成本约为每月31美元，或每小时0.04美元——相比GPU加速实例，成本降幅高达1000%！ 如前所述，若您确需即时响应的动态过滤能力，则GPU方案在此场景下确实提供了良好功能；但对于其余绝大多数用例，构建恰当的聚合瓦片集通常既能满足功能需求，又能显著降低成本。 核心问题在于：您所产出的洞察真实成本是多少？其响应速度与结果质量，是否真正匹配本用例的价值预期？同时，支撑未来各类用例的基础设施成本又该如何评估？ 本地图由CARTO CLI与CARTO AI Agents构建而成——后者是一套智能体工具，支持通过自然语言创建地图、编辑地图并执行空间分析。您可在CARTO AI Agents工具参考文档中查阅智能体所能执行的全部空间任务列表。 对大多数团队而言，将空间分析保留在数据仓库中，在成本、可扩展性与治理能力三方面均具明显优势。此举同时也为……

Analyzing and visualizing large amounts of geospatial data has sometimes seemed like a pipe dream to many analysts involved in GIS or geospatial analysis. Limitations on file sizes or infrastructure that supports spatial analysis at scale have always seemed to keep this just out of reach. In this post, we will take a look at two of these approaches: GPU accelerated analytics and cloud data warehouses with spatial capabilities. If you have seen a map or dashboard powered by GPU accelerated databases or rendering, you know that these dashboards are fast and can display massive amounts of data. They can easily store, query, and visualize millions and billions of rows of data without significant need to prepare or transform the data. While GPU accelerated analytics provide fast analytics and visualization, these services have several drawbacks, the main being price and continuous running. Before we discuss the specifics, I’ll share an analogy that can help explain this in detail. Imagine that you are deciding if you want to buy a car. You live in a city and you mostly get around with a mix of bike share, public transportation, and car share services like Uber. With your other options, you can get the service that is appropriate for you at the time you need it, and you only pay for what you need on demand. In addition if you want better service, like an upgraded car or faster pick up, you can choose to have that added cost. GPUs are like owning a car and cloud data warehouses are like using public transportation/bikes/ride share. When you are using a GPU accelerated geospatial service, the GPUs need to be running around the clock, no matter if you are using it or not. As you scale and add more data or dashboards, you may need to add more GPUs to keep performance up, and these services are expensive. Consider the cost of the cloud infrastructure alone. A GPU accelerated service running on an AWS EC2 p4d.24xlarge instance, one of the most powerful and expensive instances AWS offers, was listed at about $42.93 an hour. These estimates are also not accounting for licensing costs or other costs that may need to be accounted for. You could manually shut down and turn on the service as needed scale the service manually, or provision some autoscaling as needed. That said this extra layer of dev ops also requires time and cost. Finally, if you add other dashboards or services to this server, you may need to provision additional services as more queries against this service could cause performance issues. If you are already using a data warehouse and pushing data into a GPU accelerated platform, you are paying not only to store and query your data in your data warehouse, but also the GPU services on top of it. And for any data that lives solely within the GPU database, you lose the ability to integrate that data with other data in your data warehouse. Data warehouses are increasing in adoption and scalability, and are a great solution for managing and analyzing large amounts of geospatial data. They provide low cost and highly efficient storage and serverless or scalable compute resources with no management. This means you can run your queries and use massive computing power only for the time you need it. Today that native approach goes further with Agentic GIS. CARTO AI Agents understand natural language and run spatial tasks directly against your data warehouse, so an analyst can ask a question and get a map or an answer back without writing every query by hand. For more on why the warehouse is the right home for spatial work, see our post on why use data warehouses for geospatial analysis. Most of the GPU accelerated tools also render map tiles in raster, or PNG tile format. A more modern approach is using vector tiles that allow for far more advanced visualizations and analysis using 3D, arcs, animations, and much more. CARTO cloud native platform provides a few key differentiations from GPU accelerated analytics tools, specifically related to map tile creation and visualization. The major compromise you are making in this approach is that you are losing the capability to dynamically filter every data point and re-render the map as you do in the HEAVY AI dashboard. With some smart SQL and selective counts you can achieve some of those filters within aggregated views. The first map actually uses two tilesets, one with a point aggregation with views above zoom 17 and one that shows the raw data below 17. You can set this zoom based filtering in Builder and set up your tilesets to control for the zoom levels. The H3 map uses a new functionality to make spatial index tilesets. Once you assign an H3 or Quadbin index to the data, you can actually create the tiles with just that data point, no need for a geometry which is far faster than the other tileset creation methods. The cost is where the real savings will come. If you were to run these queries once a month and store the appropriate data this would cost roughly $31 a month or $0.04 an hour, to store all the data and tiles for this use case, or a 1000% cost saving compared to a GPU accelerated instance! As I mentioned If you need that dynamic filtering with immediate response on the fly, then the GPU solutions do provide a good functionality here, but for all other use cases creating appropriate aggregated tilesets tend to provide good functionality and cost savings. The key question to ask is what are the true cost of the insights you are producing, and do the speed and answers provide appropriate value for this use case and the infrastructure to support future use cases as well. We built this map with the CARTO CLI and CARTO AI Agents, the agentic tooling that lets you create maps, edit them, and run spatial analysis from natural language. You can see the full set of spatial tasks an agent can run in the CARTO AI Agents tools reference. For most teams, keeping spatial analysis in the data warehouse wins on cost, scale, and governance. It also sets you up for what comes next: with Agentic GIS, CARTO AI Agents can run these spatial workflows for you, directly against the data you already store in the warehouse. Want to try CARTO for yourself? Sign up for a 14-day free trial or schedule a demo with our experts. It can be. GPU databases need dedicated instances running around the clock. In one HEAVY.AI example the underlying AWS GPU instance cost about $42.93 per hour, or roughly $376,000 per year, before licensing. Running the equivalent workload as data warehouse tilesets in the example cost about $31 per month. When you need dynamic, point-level filtering with instant re-rendering across a full dataset, GPU databases do this well. For most other use cases, aggregated tilesets in a cloud data warehouse give similar interactivity at much lower cost. Matt Forrest was the Field CTO at CARTO and has worked with geospatial technology and data for over 10 years. He is passionate about all things geospatial, with a mission to help others use modern GIS and spatial SQL to improve their work and advance their careers. Public sector agencies already have Oracle's spatial engine. CARTO turns it into self-service AI Agents, giving teams instant answers without a queue. AI Agents are only as good as their understanding of your data. CARTO brings semantic models to AI Agents, with spatial context no other layer provides. Why enterprise teams move from Esri ArcGIS to cloud-native CARTO, and how an AI agent handles the migration end to end, step by step.