mirror of
https://github.com/neondatabase/neon.git
synced 2026-01-06 04:52:55 +00:00
aa4ec11af9
# Problem
The timeout-based batching adds latency to unbatchable workloads.
We can choose a short batching timeout (e.g. 10us) but that requires
high-resolution timers, which tokio doesn't have.
I thoroughly explored options to use OS timers (see
[this](https://github.com/neondatabase/neon/pull/9822) abandoned PR).
In short, it's not an attractive option because any timer implementation
adds non-trivial overheads.
# Solution
The insight is that, in the steady state of a batchable workload, the
time we spend in `get_vectored` will be hundreds of microseconds anyway.
If we prepare the next batch concurrently to `get_vectored`, we will
have a sizeable batch ready once `get_vectored` of the current batch is
done and do not need an explicit timeout.
This can be reasonably described as **pipelining of the protocol
handler**.
# Implementation
We model the sub-protocol handler for pagestream requests
(`handle_pagrequests`) as two futures that form a pipeline:
2. Batching: read requests from the connection and fill the current
batch
3. Execution: `take` the current batch, execute it using `get_vectored`,
and send the response.
The Reading and Batching stage are connected through a new type of
channel called `spsc_fold`.
See the long comment in the `handle_pagerequests_pipelined` for details.
# Changes
- Refactor `handle_pagerequests`
- separate functions for
- reading one protocol message; produces a `BatchedFeMessage` with just
one page request in it
- batching; tried to merge an incoming `BatchedFeMessage` into an
existing `BatchedFeMessage`; returns `None` on success and returns back
the incoming message in case merging isn't possible
- execution of a batched message
- unify the timeline handle acquisition & request span construction; it
now happen in the function that reads the protocol message
- Implement serial and pipelined model
- serial: what we had before any of the batching changes
- read one protocol message
- execute protocol messages
- pipelined: the design described above
- optionality for execution of the pipeline: either via concurrent
futures vs tokio tasks
- Pageserver config
- remove batching timeout field
- add ability to configure pipelining mode
- add ability to limit max batch size for pipelined configurations
(required for the rollout, cf
https://github.com/neondatabase/cloud/issues/20620 )
- ability to configure execution mode
- Tests
- remove `batch_timeout` parametrization
- rename `test_getpage_merge_smoke` to `test_throughput`
- add parametrization to test different max batch sizes and execution
moes
- rename `test_timer_precision` to `test_latency`
- rename the test case file to `test_page_service_batching.py`
- better descriptions of what the tests actually do
## On the holding The `TimelineHandle` in the pending batch
While batching, we hold the `TimelineHandle` in the pending batch.
Therefore, the timeline will not finish shutting down while we're
batching.
This is not a problem in practice because the concurrently ongoing
`get_vectored` call will fail quickly with an error indicating that the
timeline is shutting down.
This results in the Execution stage returning a `QueryError::Shutdown`,
which causes the pipeline / entire page service connection to shut down.
This drops all references to the
`Arc<Mutex<Option<Box<BatchedFeMessage>>>>` object, thereby dropping the
contained `TimelineHandle`s.
- => fixes https://github.com/neondatabase/neon/issues/9850
# Performance
Local run of the benchmarks, results in [this empty
commit](1cf5b1463f)
in the PR branch.
Key take-aways:
* `concurrent-futures` and `tasks` deliver identical `batching_factor`
* tail latency impact unknown, cf
https://github.com/neondatabase/neon/issues/9837
* `concurrent-futures` has higher throughput than `tasks` in all
workloads (=lower `time` metric)
* In unbatchable workloads, `concurrent-futures` has 5% higher
`CPU-per-throughput` than that of `tasks`, and 15% higher than that of
`serial`.
* In batchable-32 workload, `concurrent-futures` has 8% lower
`CPU-per-throughput` than that of `tasks` (comparison to tput of
`serial` is irrelevant)
* in unbatchable workloads, mean and tail latencies of
`concurrent-futures` is practically identical to `serial`, whereas
`tasks` adds 20-30us of overhead
Overall, `concurrent-futures` seems like a slightly more attractive
choice.
# Rollout
This change is disabled-by-default.
Rollout plan:
- https://github.com/neondatabase/cloud/issues/20620
# Refs
- epic: https://github.com/neondatabase/neon/issues/9376
- this sub-task: https://github.com/neondatabase/neon/issues/9377
- the abandoned attempt to improve batching timeout resolution:
https://github.com/neondatabase/neon/pull/9820
- closes https://github.com/neondatabase/neon/issues/9850
- fixes https://github.com/neondatabase/neon/issues/9835
320 lines
11 KiB
TOML
320 lines
11 KiB
TOML
[workspace]
|
|
resolver = "2"
|
|
members = [
|
|
"compute_tools",
|
|
"control_plane",
|
|
"control_plane/storcon_cli",
|
|
"pageserver",
|
|
"pageserver/compaction",
|
|
"pageserver/ctl",
|
|
"pageserver/client",
|
|
"pageserver/pagebench",
|
|
"proxy",
|
|
"safekeeper",
|
|
"storage_broker",
|
|
"storage_controller",
|
|
"storage_controller/client",
|
|
"storage_scrubber",
|
|
"workspace_hack",
|
|
"libs/compute_api",
|
|
"libs/pageserver_api",
|
|
"libs/postgres_ffi",
|
|
"libs/safekeeper_api",
|
|
"libs/desim",
|
|
"libs/utils",
|
|
"libs/consumption_metrics",
|
|
"libs/postgres_backend",
|
|
"libs/pq_proto",
|
|
"libs/tenant_size_model",
|
|
"libs/metrics",
|
|
"libs/postgres_connection",
|
|
"libs/remote_storage",
|
|
"libs/tracing-utils",
|
|
"libs/postgres_ffi/wal_craft",
|
|
"libs/vm_monitor",
|
|
"libs/walproposer",
|
|
"libs/wal_decoder",
|
|
"libs/postgres_initdb",
|
|
"libs/proxy/postgres-protocol2",
|
|
"libs/proxy/postgres-types2",
|
|
"libs/proxy/tokio-postgres2",
|
|
]
|
|
|
|
[workspace.package]
|
|
edition = "2021"
|
|
license = "Apache-2.0"
|
|
|
|
## All dependency versions, used in the project
|
|
[workspace.dependencies]
|
|
ahash = "0.8"
|
|
anyhow = { version = "1.0", features = ["backtrace"] }
|
|
arc-swap = "1.6"
|
|
async-compression = { version = "0.4.0", features = ["tokio", "gzip", "zstd"] }
|
|
atomic-take = "1.1.0"
|
|
azure_core = { version = "0.19", default-features = false, features = ["enable_reqwest_rustls", "hmac_rust"] }
|
|
azure_identity = { version = "0.19", default-features = false, features = ["enable_reqwest_rustls"] }
|
|
azure_storage = { version = "0.19", default-features = false, features = ["enable_reqwest_rustls"] }
|
|
azure_storage_blobs = { version = "0.19", default-features = false, features = ["enable_reqwest_rustls"] }
|
|
flate2 = "1.0.26"
|
|
async-stream = "0.3"
|
|
async-trait = "0.1"
|
|
aws-config = { version = "1.5", default-features = false, features=["rustls", "sso"] }
|
|
aws-sdk-s3 = "1.52"
|
|
aws-sdk-iam = "1.46.0"
|
|
aws-sdk-kms = "1.47.0"
|
|
aws-smithy-async = { version = "1.2.1", default-features = false, features=["rt-tokio"] }
|
|
aws-smithy-types = "1.2"
|
|
aws-credential-types = "1.2.0"
|
|
aws-sigv4 = { version = "1.2", features = ["sign-http"] }
|
|
aws-types = "1.3"
|
|
axum = { version = "0.7.5", features = ["ws"] }
|
|
base64 = "0.13.0"
|
|
bincode = "1.3"
|
|
bindgen = "0.70"
|
|
bit_field = "0.10.2"
|
|
bstr = "1.0"
|
|
byteorder = "1.4"
|
|
bytes = "1.0"
|
|
camino = "1.1.6"
|
|
cfg-if = "1.0.0"
|
|
chrono = { version = "0.4", default-features = false, features = ["clock"] }
|
|
clap = { version = "4.0", features = ["derive", "env"] }
|
|
comfy-table = "7.1"
|
|
const_format = "0.2"
|
|
crc32c = "0.6"
|
|
dashmap = { version = "5.5.0", features = ["raw-api"] }
|
|
diatomic-waker = { version = "0.2.3" }
|
|
either = "1.8"
|
|
enum-map = "2.4.2"
|
|
enumset = "1.0.12"
|
|
fail = "0.5.0"
|
|
fallible-iterator = "0.2"
|
|
framed-websockets = { version = "0.1.0", git = "https://github.com/neondatabase/framed-websockets" }
|
|
futures = "0.3"
|
|
futures-core = "0.3"
|
|
futures-util = "0.3"
|
|
git-version = "0.3"
|
|
hashbrown = "0.14"
|
|
hashlink = "0.9.1"
|
|
hdrhistogram = "7.5.2"
|
|
hex = "0.4"
|
|
hex-literal = "0.4"
|
|
hmac = "0.12.1"
|
|
hostname = "0.4"
|
|
http = {version = "1.1.0", features = ["std"]}
|
|
http-types = { version = "2", default-features = false }
|
|
http-body-util = "0.1.2"
|
|
humantime = "2.1"
|
|
humantime-serde = "1.1.1"
|
|
hyper0 = { package = "hyper", version = "0.14" }
|
|
hyper = "1.4"
|
|
hyper-util = "0.1"
|
|
tokio-tungstenite = "0.21.0"
|
|
indexmap = "2"
|
|
indoc = "2"
|
|
ipnet = "2.10.0"
|
|
itertools = "0.10"
|
|
itoa = "1.0.11"
|
|
jsonwebtoken = "9"
|
|
lasso = "0.7"
|
|
libc = "0.2"
|
|
md5 = "0.7.0"
|
|
measured = { version = "0.0.22", features=["lasso"] }
|
|
measured-process = { version = "0.0.22" }
|
|
memoffset = "0.9"
|
|
nix = { version = "0.27", features = ["dir", "fs", "process", "socket", "signal", "poll"] }
|
|
notify = "6.0.0"
|
|
num_cpus = "1.15"
|
|
num-traits = "0.2.15"
|
|
once_cell = "1.13"
|
|
opentelemetry = "0.24"
|
|
opentelemetry_sdk = "0.24"
|
|
opentelemetry-otlp = { version = "0.17", default-features=false, features = ["http-proto", "trace", "http", "reqwest-client"] }
|
|
opentelemetry-semantic-conventions = "0.16"
|
|
parking_lot = "0.12"
|
|
parquet = { version = "53", default-features = false, features = ["zstd"] }
|
|
parquet_derive = "53"
|
|
pbkdf2 = { version = "0.12.1", features = ["simple", "std"] }
|
|
pin-project-lite = "0.2"
|
|
pprof = { version = "0.14", features = ["criterion", "flamegraph", "protobuf", "protobuf-codec"] }
|
|
procfs = "0.16"
|
|
prometheus = {version = "0.13", default-features=false, features = ["process"]} # removes protobuf dependency
|
|
prost = "0.13"
|
|
rand = "0.8"
|
|
redis = { version = "0.25.2", features = ["tokio-rustls-comp", "keep-alive"] }
|
|
regex = "1.10.2"
|
|
reqwest = { version = "0.12", default-features = false, features = ["rustls-tls"] }
|
|
reqwest-tracing = { version = "0.5", features = ["opentelemetry_0_24"] }
|
|
reqwest-middleware = "0.3.0"
|
|
reqwest-retry = "0.5"
|
|
routerify = "3"
|
|
rpds = "0.13"
|
|
rustc-hash = "1.1.0"
|
|
rustls = { version = "0.23.16", default-features = false }
|
|
rustls-pemfile = "2"
|
|
scopeguard = "1.1"
|
|
sysinfo = "0.29.2"
|
|
sd-notify = "0.4.1"
|
|
send-future = "0.1.0"
|
|
sentry = { version = "0.32", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
|
|
serde = { version = "1.0", features = ["derive"] }
|
|
serde_json = "1"
|
|
serde_path_to_error = "0.1"
|
|
serde_with = { version = "2.0", features = [ "base64" ] }
|
|
serde_assert = "0.5.0"
|
|
sha2 = "0.10.2"
|
|
signal-hook = "0.3"
|
|
smallvec = "1.11"
|
|
smol_str = { version = "0.2.0", features = ["serde"] }
|
|
socket2 = "0.5"
|
|
strum = "0.26"
|
|
strum_macros = "0.26"
|
|
"subtle" = "2.5.0"
|
|
svg_fmt = "0.4.3"
|
|
sync_wrapper = "0.1.2"
|
|
tar = "0.4"
|
|
test-context = "0.3"
|
|
thiserror = "1.0"
|
|
tikv-jemallocator = { version = "0.6", features = ["stats"] }
|
|
tikv-jemalloc-ctl = { version = "0.6", features = ["stats"] }
|
|
tokio = { version = "1.17", features = ["macros"] }
|
|
tokio-epoll-uring = { git = "https://github.com/neondatabase/tokio-epoll-uring.git" , branch = "main" }
|
|
tokio-io-timeout = "1.2.0"
|
|
tokio-postgres-rustls = "0.12.0"
|
|
tokio-rustls = { version = "0.26.0", default-features = false, features = ["tls12", "ring"]}
|
|
tokio-stream = "0.1"
|
|
tokio-tar = "0.3"
|
|
tokio-util = { version = "0.7.10", features = ["io", "rt"] }
|
|
toml = "0.8"
|
|
toml_edit = "0.22"
|
|
tonic = {version = "0.12.3", features = ["tls", "tls-roots"]}
|
|
tower-service = "0.3.2"
|
|
tracing = "0.1"
|
|
tracing-error = "0.2"
|
|
tracing-opentelemetry = "0.25"
|
|
tracing-subscriber = { version = "0.3", default-features = false, features = ["smallvec", "fmt", "tracing-log", "std", "env-filter", "json"] }
|
|
try-lock = "0.2.5"
|
|
twox-hash = { version = "1.6.3", default-features = false }
|
|
typed-json = "0.1"
|
|
url = "2.2"
|
|
urlencoding = "2.1"
|
|
uuid = { version = "1.6.1", features = ["v4", "v7", "serde"] }
|
|
walkdir = "2.3.2"
|
|
rustls-native-certs = "0.8"
|
|
x509-parser = "0.16"
|
|
whoami = "1.5.1"
|
|
zerocopy = { version = "0.7", features = ["derive"] }
|
|
|
|
## TODO replace this with tracing
|
|
env_logger = "0.10"
|
|
log = "0.4"
|
|
|
|
## Libraries from neondatabase/ git forks, ideally with changes to be upstreamed
|
|
postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
|
|
postgres-protocol = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
|
|
postgres-types = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
|
|
tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
|
|
|
|
## Local libraries
|
|
compute_api = { version = "0.1", path = "./libs/compute_api/" }
|
|
consumption_metrics = { version = "0.1", path = "./libs/consumption_metrics/" }
|
|
metrics = { version = "0.1", path = "./libs/metrics/" }
|
|
pageserver = { path = "./pageserver" }
|
|
pageserver_api = { version = "0.1", path = "./libs/pageserver_api/" }
|
|
pageserver_client = { path = "./pageserver/client" }
|
|
pageserver_compaction = { version = "0.1", path = "./pageserver/compaction/" }
|
|
postgres_backend = { version = "0.1", path = "./libs/postgres_backend/" }
|
|
postgres_connection = { version = "0.1", path = "./libs/postgres_connection/" }
|
|
postgres_ffi = { version = "0.1", path = "./libs/postgres_ffi/" }
|
|
postgres_initdb = { path = "./libs/postgres_initdb" }
|
|
pq_proto = { version = "0.1", path = "./libs/pq_proto/" }
|
|
remote_storage = { version = "0.1", path = "./libs/remote_storage/" }
|
|
safekeeper_api = { version = "0.1", path = "./libs/safekeeper_api" }
|
|
desim = { version = "0.1", path = "./libs/desim" }
|
|
storage_broker = { version = "0.1", path = "./storage_broker/" } # Note: main broker code is inside the binary crate, so linking with the library shouldn't be heavy.
|
|
storage_controller_client = { path = "./storage_controller/client" }
|
|
tenant_size_model = { version = "0.1", path = "./libs/tenant_size_model/" }
|
|
tracing-utils = { version = "0.1", path = "./libs/tracing-utils/" }
|
|
utils = { version = "0.1", path = "./libs/utils/" }
|
|
vm_monitor = { version = "0.1", path = "./libs/vm_monitor/" }
|
|
walproposer = { version = "0.1", path = "./libs/walproposer/" }
|
|
wal_decoder = { version = "0.1", path = "./libs/wal_decoder" }
|
|
|
|
## Common library dependency
|
|
workspace_hack = { version = "0.1", path = "./workspace_hack/" }
|
|
|
|
## Build dependencies
|
|
criterion = "0.5.1"
|
|
rcgen = "0.13"
|
|
rstest = "0.18"
|
|
camino-tempfile = "1.0.2"
|
|
tonic-build = "0.12"
|
|
|
|
[patch.crates-io]
|
|
|
|
# Needed to get `tokio-postgres-rustls` to depend on our fork.
|
|
tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", branch = "neon" }
|
|
|
|
################# Binary contents sections
|
|
|
|
[profile.release]
|
|
# This is useful for profiling and, to some extent, debug.
|
|
# Besides, debug info should not affect the performance.
|
|
debug = true
|
|
|
|
# disable debug symbols for all packages except this one to decrease binaries size
|
|
[profile.release.package."*"]
|
|
debug = false
|
|
|
|
[profile.release-line-debug]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
[profile.release-line-debug-lto]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
lto = true
|
|
|
|
[profile.release-line-debug-size]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
opt-level = "s"
|
|
[profile.release-line-debug-zize]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
opt-level = "z"
|
|
[profile.release-line-debug-size-lto]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
opt-level = "s"
|
|
lto = true
|
|
[profile.release-line-debug-zize-lto]
|
|
inherits = "release"
|
|
debug = 1 # true = 2 = all symbols, 1 = line only
|
|
opt-level = "z"
|
|
lto = true
|
|
|
|
[profile.release-no-debug]
|
|
inherits = "release"
|
|
debug = false # true = 2 = all symbols, 1 = line only
|
|
|
|
[profile.release-no-debug-size]
|
|
inherits = "release"
|
|
debug = false # true = 2 = all symbols, 1 = line only
|
|
opt-level = "s"
|
|
[profile.release-no-debug-zize]
|
|
inherits = "release"
|
|
debug = false # true = 2 = all symbols, 1 = line only
|
|
opt-level = "z"
|
|
|
|
[profile.release-no-debug-size-lto]
|
|
inherits = "release"
|
|
debug = false # true = 2 = all symbols, 1 = line only
|
|
opt-level = "s"
|
|
lto = true
|
|
|
|
[profile.release-no-debug-zize-lto]
|
|
inherits = "release"
|
|
debug = false # true = 2 = all symbols, 1 = line only
|
|
opt-level = "z"
|
|
lto = true
|