你的服务器够快吗？一文说清如何衡量服务器的性能？

# --- 单线程函数 (serial_func) ---
# ... Seurat 标准流程步骤 ...
obj.markers <- FindAllMarkers(obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)
# ---

# --- 并行处理函数 (parallel_func) ---
library(future)
plan("multisession", workers = 8) # <--- 可配置参数：工作核心数
options(future.globals.maxSize = 128 * 1024 * 1024^2) # <--- 可配置参数：每个核心允许的全局变量大小 (内存相关)
# ... Seurat 标准流程步骤 ...
# 并行计算标记基因
obj.markers <- future({
  FindAllMarkers(obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)
}, globals = list(obj = obj, FindAllMarkers = FindAllMarkers))
result <- value(obj.markers) # 获取并行结果
# ---

# --- 计时 ---
system.time(serial_func())
system.time(parallel_func())
# ---

library(logging)
logReset()
basicConfig(level='INFO')
addHandler(writeToFile, file="./demo6.log", level='INFO')

# 单线程处理函数
serial_func <- function() {
  library(Seurat)
  obj <- readRDS("Parkinson.rds")
  DefaultAssay(obj) <- "RNA"
  obj[["percent.mt"]] <- PercentageFeatureSet(obj, pattern = "^MT-", assay = "RNA")
  obj <- NormalizeData(obj, normalization.method = "LogNormalize", scale.factor = 10000)
  obj <- FindVariableFeatures(obj, selection.method = "vst", nfeatures = 2000)
  obj <- ScaleData(obj, features = VariableFeatures(obj))  # 只缩放高变基因
  obj <- RunPCA(obj, features = VariableFeatures(obj))
  obj <- FindNeighbors(obj, dims = 1:10)
  obj <- FindClusters(obj, resolution = 0.5)
  obj <- RunUMAP(obj, dims = 1:10)
  obj.markers <- FindAllMarkers(obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)
return(obj.markers)
}

# 优化后的并行处理函数
parallel_func <- function() {
  library(future)
  library(Seurat)

# 设置并行后端 - 考虑到内存限制，使用2个worker
  plan("multisession", workers = 2)

# 正确设置内存限制 (128GB)
  options(future.globals.maxSize = 128 * 1024^2 * 1024)

# 设置随机数种子以避免警告
  set.seed(42)

# 启用Seurat的并行计算
  options(future.seed = TRUE)

  obj <- readRDS("Parkinson.rds")
  DefaultAssay(obj) <- "RNA"
  obj[["percent.mt"]] <- PercentageFeatureSet(obj, pattern = "^MT-", assay = "RNA")

# 使用并行计算进行数据标准化
  obj <- NormalizeData(obj, normalization.method = "LogNormalize", scale.factor = 10000)
  obj <- FindVariableFeatures(obj, selection.method = "vst", nfeatures = 2000)

# 使用并行计算进行数据缩放
  obj <- ScaleData(obj, features = VariableFeatures(obj))

# 使用并行计算进行PCA
  obj <- RunPCA(obj, features = VariableFeatures(obj))

# 使用并行计算进行聚类
  obj <- FindNeighbors(obj, dims = 1:10)
  obj <- FindClusters(obj, resolution = 0.5)

# 使用并行计算进行UMAP降维
  obj <- RunUMAP(obj, dims = 1:10)

# 并行计算标记基因 - 这一步是最耗时的
  obj.markers <- FindAllMarkers(obj, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)

return(obj.markers)
}

# 记录执行时间
loginfo("开始单线程处理...")
time_serial <- system.time(serial_results <- serial_func())
loginfo("单线程完成。耗时: user=%s, system=%s, elapsed=%s", time_serial['user.self'], time_serial['sys.self'], time_serial['elapsed'])

loginfo("开始多线程处理...")
time_parallel <- system.time(parallel_results <- parallel_func())
loginfo("多线程完成。耗时: user=%s, system=%s, elapsed=%s", time_parallel['user.self'], time_parallel['sys.self'], time_parallel['elapsed'])

# 比较两种方法的速度提升
speedup <- time_serial['elapsed'] / time_parallel['elapsed']
loginfo("速度提升: %.2f 倍", speedup)

library(logging)
logReset()
basicConfig(level = "INFO")
# 设置输出日志到文件
addHandler(writeToFile, file = "~/demo.log", level = "INFO")
rm(list = ls())
set.seed(123)
# 设置矩阵的行数
n <- 5000
# 生成一个矩阵
value <- rnorm(n * n, 10, 3)
mat <- matrix(value, n, n)
result1 <- system.time({
        # 矩阵求逆
        ainv <- solve(mat)})
result1
loginfo("求逆耗时 %s", result1)
result2 <- system.time({
        # 矩阵相乘
        re <- mat %*% t(mat)})
result2

loginfo("相乘耗时 %s", result2)