Skip to content

Commit ace0199

Browse files
committed
doc: update performance details for snapshots
Signed-off-by: Diana Popa <dpopa@amazon.com>
1 parent b2d36f7 commit ace0199

File tree

1 file changed

+21
-13
lines changed

1 file changed

+21
-13
lines changed

docs/snapshotting/snapshot-support.md

Lines changed: 21 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -93,26 +93,34 @@ snapshot files by implementing authentication and encryption schemes while
9393
managing their lifecycle or moving them across the trust boundary, like for
9494
example when provisioning them from a respository to a host over the network.
9595

96-
Firecracker is optimized for fast load/resume and it's designed to do some very basic
97-
sanity checks only on the vm state file. It only verifies integrity using a 64
98-
bit CRC value embedded in the vm state file, but this is only as a partial
99-
measure to protect against accidental corruption, as the disk files and memory
100-
file need to be secured as well. It is important to note that CRC computation
101-
is validated before trying to load the snapshot. Should it encounter failure,
102-
an error will be shown to the user and the Firecracker process will be terminated.
96+
Firecracker is optimized for fast load/resume, and it's designed to do some
97+
very basic sanity checks only on the vm state file. It only verifies integrity
98+
using a 64-bit CRC value embedded in the vm state file, but this is only
99+
as a partial measure to protect against accidental corruption, as the disk
100+
files and memory file need to be secured as well. It is important to note that
101+
CRC computation is validated before trying to load the snapshot. Should it
102+
encounter failure, an error will be shown to the user and the Firecracker
103+
process will be terminated.
103104

104105
### Performance
105106

106107
The Firecracker snapshot create/resume performance depends on the memory size,
107-
vCPU count and emulated devices count. The Firecracker CI runs snapshots tests
108-
on AWS **m5d.metal** instances for Intel and on AWS **m6g.metal** for ARM.
109-
The baseline for snapshot resume latency target on Intel is under **8ms** with
110-
5ms p90, and on ARM is under **3ms** for a microVM with the following specs:
111-
2vCPU/512MB/1 block/1 net device.
108+
vCPU count and emulated devices count.
109+
The Firecracker CI runs snapshots tests on:
110+
111+
- AWS **m5d.metal** and **m6i.metal** instances for Intel
112+
- AWS **m6g.metal** for ARM
113+
- AWS **m6a.metal** for AMD
114+
115+
We are running nightly performance tests for all the enumerated platforms on
116+
all supported kernel versions.
117+
The baselines can be found in their [respective config file](../../tests/integration_tests/performance/configs/).
112118

113119
### Known issues
114120

115-
- High snapshot latency on 5.4+ host kernels - [#2129](https://github.com/firecracker-microvm/firecracker/issues/2129)
121+
- High snapshot latency on 5.4+ host kernels due to cgroups V1. We
122+
- strongly recommend to deploy snapshots on cgroups V2 enabled hosts for the
123+
- implied kernel versions - [related issue](https://github.com/firecracker-microvm/firecracker/issues/2129).
116124
- Guest network connectivity is not guaranteed to be preserved after resume.
117125
For recommendations related to guest network connectivity for clones please
118126
see [Network connectivity for clones](network-for-clones.md).

0 commit comments

Comments
 (0)