README
¶
DaisyNFS evaluation
The AWS evaluation setup is documented in aws/README.md. These instructions document a more manual setup.
The evaluation scripts use some environment variables that point to a few repositories:
export GO_NFSD_PATH=~/go-nfsd
export DAISY_NFSD_PATH=~/daisy-nfsd
export XV6_PATH=~/xv6-public
export LTP_PATH=~/ltp
You'll need to clone mit-pdos/go-nfsd, mit-pdos/xv6-public, and linux-test-project/ltp (this last one is only needed to run the stress tests).
This repo is mit-pdos/daisy-nfsd.
These instructions assume you've compiled the evaluation driver with go build ./cmd/daisy-eval.
smallfile, largefile, and app benchmarks
Run daisy-eval -i eval/data bench. Then ./eval/eval.py -i eval/data bench
will produce a file eval/data/bench.data.
smallfile scalability on a disk
Run daisy-eval -i eval/data bench. Then ./eval/eval.py -i eval/data scale
will produce files for each system in data/{daisy-nfsd,go-nfsd,linux}.data.
The daisy-eval driver has an argument to set the disk file.
Plotting
You can run ./plot.sh to run the Python post-processing and gnuplot all at once.
stress tests
Run ./tests.sh to run the fsstress and fsx-linux tests. You'll need to clone
ltp and compile it; running ./tests.sh --help will give you the right commands.
The default number of iterations runs each suite for about 10 seconds. To run
longer, run something like ./tests.sh 10, which will scale the default
iteration counts by 10.
Documentation
¶
Index ¶
- Variables
- func PrepareBenchmarks()
- func WriteObservation(w io.Writer, o Observation)
- type Benchmark
- func AppBench() Benchmark
- func BenchSuite(smallfileDuration string) []Benchmark
- func ExtendedBenchSuite(smallfileDuration string, par int) []Benchmark
- func LargefileBench(fileSizeMb int) Benchmark
- func ScaleSuite(benchtime string, threads int) []Benchmark
- func SmallfileBench(benchtime string, threads int) Benchmark
- type BenchmarkSuite
- type Fs
- type KeyValue
- type KeyValuePair
- type Observation
- type Workload
Constants ¶
This section is empty.
Variables ¶
var LargefileSuite = []Benchmark{ LargefileBench(300), }
Functions ¶
func PrepareBenchmarks ¶
func PrepareBenchmarks()
func WriteObservation ¶
func WriteObservation(w io.Writer, o Observation)
Types ¶
type Benchmark ¶
type Benchmark struct {
// Config has configuration related to the benchmark workload under test
// "bench" is a map with benchmark options
Config KeyValue
// contains filtered or unexported fields
}
func BenchSuite ¶
func ExtendedBenchSuite ¶
func LargefileBench ¶
func ScaleSuite ¶
func SmallfileBench ¶
func (Benchmark) ParseOutput ¶
func (b Benchmark) ParseOutput(lines []string) []Observation
type BenchmarkSuite ¶
func (*BenchmarkSuite) Workloads ¶
func (bs *BenchmarkSuite) Workloads() []Workload
type KeyValue ¶
type KeyValue map[string]interface{}
KeyValue is a generic set of key-value pairs
expect values to be string, float64, or bool (or recursively another KeyValue)
func LinuxDurabilityFilesystems ¶
LinuxDurabilityFilesystems returns many Linux filesystems, varying durability options
func (KeyValue) Extend ¶
Extend adds all key-value pairs from kv2 to kv
modifies kv in-place and returns kv (for chaining)
func (KeyValue) Pairs ¶
func (kv KeyValue) Pairs() []KeyValuePair
Pairs returns the key-value pairs in kv, sorted by key
type KeyValuePair ¶
type KeyValuePair struct {
Key string
Val interface{}
}
type Observation ¶
func ReadObservation ¶
func ReadObservation(r io.Reader) (o Observation, err error)
ReadObservation gets the next observation in r
Source Files