评估

Prisma Accelerate 通过先进的连接池和全局边缘缓存优化数据库交互。其连接池可在 16 个区域使用，并帮助应用程序根据需求进行负载均衡和扩展数据库请求。

考虑到以上信息，我们建议在高负载下评估 Accelerate，以观察其性能表现。

Accelerate 的连接池如何优化高负载下的性能

Prisma Accelerate 采用动态的无服务器连接池基础设施。当发出请求时，会在配置 Prisma Accelerate 时指定的区域中为项目快速配置连接池。此连接池保持活动状态，在重用已建立的数据库连接的同时，处理更多额外的请求。连接池会在一段时间不活动后断开连接，因此务必使用持续的流量来评估 Prisma Accelerate。

主要优势

• 优化的查询性能： 无服务器连接池会适应查询负载，确保在峰值需求期间高效管理数据库连接。

  Prisma Accelerate 的连接池无法提高数据库中查询的性能。如果查询性能存在问题，我们建议优化 Prisma 查询、应用索引或利用 Accelerate 的边缘缓存。

• 最大化连接重用： 执行持续大量的查询有助于保持 Accelerate 连接池的活动实例。这会增加连接重用，确保后续查询的响应时间更快。

通过理解和利用这种机制，您可以确保数据库查询在大规模下持续高效地执行。

评估 Prisma Accelerate 连接池性能

以下您将找到一个如何使用示例模型评估 Prisma Accelerate 的示例

model Notes {
  id        Int       @id @default(autoincrement())
  title     String
  createdAt DateTime  @default(now())
  updatedAt DateTime? @updatedAt
}

import { PrismaClient } from '@prisma/client'
import { withAccelerate } from '@prisma/extension-accelerate'

const prisma = new PrismaClient().$extends(withAccelerate())

function calculateStatistics(numbers: number[]): {
  average: number
  p50: number
  p75: number
  p99: number
} {
  if (numbers.length === 0) {
    throw new Error('The input array is empty.')
  }

  // Sort the array in ascending order
  numbers.sort((a, b) => a - b)

  const sum = numbers.reduce((acc, num) => acc + num, 0)
  const count = numbers.length

  const average = sum / count
  const p50 = getPercentile(numbers, 50)
  const p75 = getPercentile(numbers, 75)
  const p99 = getPercentile(numbers, 99)

  return { average, p50, p75, p99 }
}

function getPercentile(numbers: number[], percentile: number): number {
  if (percentile <= 0 || percentile >= 100) {
    throw new Error('Percentile must be between 0 and 100.')
  }

  const index = (percentile / 100) * (numbers.length - 1)
  if (Number.isInteger(index)) {
    // If the index is an integer, return the corresponding value
    return numbers[index]
  } else {
    // If the index is not an integer, interpolate between two adjacent values
    const lowerIndex = Math.floor(index)
    const upperIndex = Math.ceil(index)
    const lowerValue = numbers[lowerIndex]
    const upperValue = numbers[upperIndex]
    const interpolationFactor = index - lowerIndex
    return lowerValue + (upperValue - lowerValue) * interpolationFactor
  }
}

async function main() {
  const timings = []

  // fire a query before going to the loop
  await prisma.notes.findMany({
    take: 20,
  })

  // we recommend evaluationg Prisma Accelerate with a large loop
  const LOOP_LENGTH = 10000

  for (let i = 0; i < LOOP_LENGTH; i++) {
    const start = Date.now()
    await prisma.notes.findMany({
      take: 20,
    })

    timings.push(Date.now() - start)
  }

  const statistics = calculateStatistics(timings)
  console.log('Average:', statistics.average)
  console.log('P50:', statistics.p50)
  console.log('P75:', statistics.p75)
  console.log('P99:', statistics.p99)
}

main()
  .then(async () => {
    await prisma.$disconnect()
  })
  .catch((e) => {
    await prisma.$disconnect()
    process.exit(1)
  })

评估 Prisma Accelerate 缓存性能

Prisma Accelerate 的边缘缓存也针对大量查询进行了优化。缓存会自动针对重复查询进行优化。因此，缓存命中率会随着查询频率的增加而提高。将查询结果添加到缓存也是非阻塞的，因此短时间的查询突发可能无法利用缓存或持续的负载。

要评估 Accelerate 的边缘缓存，您可以使用以下内容修改上述脚本

import { PrismaClient } from '@prisma/client'
import { withAccelerate } from '@prisma/extension-accelerate'

const prisma = new PrismaClient().$extends(withAccelerate())

function calculateStatistics(numbers: number[]): {
  average: number
  p50: number
  p75: number
  p99: number
} {
  if (numbers.length === 0) {
    throw new Error('The input array is empty.')
  }

  // Sort the array in ascending order
  numbers.sort((a, b) => a - b)

  const sum = numbers.reduce((acc, num) => acc + num, 0)
  const count = numbers.length

  const average = sum / count
  const p50 = getPercentile(numbers, 50)
  const p75 = getPercentile(numbers, 75)
  const p99 = getPercentile(numbers, 99)

  return { average, p50, p75, p99 }
}

function getPercentile(numbers: number[], percentile: number): number {
  if (percentile <= 0 || percentile >= 100) {
    throw new Error('Percentile must be between 0 and 100.')
  }

  const index = (percentile / 100) * (numbers.length - 1)
  if (Number.isInteger(index)) {
    // If the index is an integer, return the corresponding value
    return numbers[index]
  } else {
    // If the index is not an integer, interpolate between two adjacent values
    const lowerIndex = Math.floor(index)
    const upperIndex = Math.ceil(index)
    const lowerValue = numbers[lowerIndex]
    const upperValue = numbers[upperIndex]
    const interpolationFactor = index - lowerIndex
    return lowerValue + (upperValue - lowerValue) * interpolationFactor
  }
}

async function main() {
  const timings = []

  // fire a query before going to the loop
  await prisma.notes.findMany({
    take: 20,
    cacheStrategy: {
      ttl: 30,
    },
  })

  // we recommend evaluating Prisma Accelerate with a large loop
  const LOOP_LENGTH = 10000

  for (let i = 0; i < LOOP_LENGTH; i++) {
    const start = Date.now()
    await prisma.notes.findMany({
      take: 20,
      cacheStrategy: {
        ttl: 30,
      },
    })

    timings.push(Date.now() - start)
  }

  const statistics = calculateStatistics(timings)
  console.log('Average:', statistics.average)
  console.log('P50:', statistics.p50)
  console.log('P75:', statistics.p75)
  console.log('P99:', statistics.p99)
}

main()
  .then(async () => {
    await prisma.$disconnect()
  })
  .catch((e) => {
    await prisma.$disconnect()
    process.exit(1)
  })