ovni/doc/runtime/trace_spec.md

211 lines
6.8 KiB
Markdown

# Trace specification
!!! Important
This document refers to the trace specification for
the version 1
The ovni instrumentation library stores the information collected in a
trace following the specification of this document.
The complete trace is stored in a top-level directory named `ovni`.
Inside this directory you will find the loom directories with the prefix
`loom.`. The name of the loom is built from the `loom` parameter of
`ovni_proc_init()`, prefixing it with `loom.`.
Each loom directory contains one directory per process of that loom. The
name is composed of the `proc.` prefix and the PID of the process
specified in the `pid` argument to `ovni_proc_init()`.
Each process directory contains:
- The metadata file `metadata.json`.
- The thread streams with prefix `thread.`.
## Process metadata
The metadata file contains important information about the trace that is
invariant during the complete execution, and generally is required to be
available prior to processing the events in the trace.
The metadata is stored in the JSON file `metadata.json` inside each
process directory and contains the following keys:
- `version`: a number specifying the version of the metadata format.
- `model_version`: a string with the version of each model supported by
the emulator.
- `app_id`: the application ID, used to distinguish between applications
running on the same loom.
- `rank`: the rank of the MPI process (optional).
- `nranks`: number of total MPI processes (optional).
- `cpus`: the array of $`N_c`$ CPUs available in the loom. Only one
process in the loom must contain this mandatory key. Each element is a
dictionary with the keys:
- `index`: containing the logical CPU index from 0 to $`N_c - 1`$.
- `phyid`: the number of the CPU as given by the operating system
(which can exceed $`N_c`$).
Here is an example of the `metadata.json` file:
```
{
"version": 1,
"model_version": "O1 V1 T1 M1 D1 K1",
"app_id": 1,
"rank": 0,
"nranks": 4,
"cpus": [
{
"index": 0,
"phyid": 0
},
{
"index": 1,
"phyid": 1
},
{
"index": 2,
"phyid": 2
},
{
"index": 3,
"phyid": 3
}
]
}
```
## Thread streams
Streams are a binary files that contains a succession of events with
monotonically increasing clock values. Streams have a small header and
the variable size events just after the header.
The header contains the magic 4 bytes of "ovni" and a version number of
4 bytes too. Here is a figure of the data stored in disk:
![Stream](fig/stream.svg)
Similarly, events have a fixed size header followed by an optional
payload of varying size. The header has the following information:
- Event flags
- Payload size in a special format
- Model, category and value codes
- Time in nanoseconds
The event size can vary depending on the data stored in the payload. The
payload size is specified using 4 bits, with the value `0x0` for no
payload, or with value $`v`$ for $`v + 1`$ bytes of payload. This allows
us to use 16 bytes of payload with value `0xf` at the cost of
sacrificing payloads of one byte.
There are two types of events, depending of the size needed for the
payload:
- Normal events: with a payload up to 16 bytes
- Jumbo events: with a payload up to $`2^{32}`$ bytes
## Normal events
The normal events are composed of:
- 4 bits of flags
- 4 bits of payload size
- 3 bytes for the MCV
- 8 bytes for the clock
- 0 to 16 bytes of payload
Here is an example of a normal event without payload, a total of 12
bytes:
```
00 4f 48 65 01 c5 cf 1d 96 d0 12 00 |.OHe........|
```
And in the following figure you can see every field annotated:
![Normal event without payload](fig/event-normal.svg)
Another example of a normal event with 16 bytes of payload, a total of
28 bytes:
```
0f 4f 48 78 58 c1 b0 b5 95 43 11 00 00 00 00 00 |.OHxX....C......|
ff ff ff ff 00 00 00 00 00 00 00 00 |............|
```
In the following figure you can see each field annotated:
![Normal event with payload content](fig/event-normal-payload.svg)
## Jumbo events
The jumbo events are just like normal events but they can hold large
data. The size of the jumbo data is stored as a 32 bits integer as a
normal payload, and the jumbo data just follows the event.
- 4 bits of flags
- 4 bits of payload size (always 4 with value 0x3)
- 3 bytes for the MCV
- 8 bytes for the clock
- 4 bytes of payload with the size of the jumbo data
- 0 to 2^32 bytes of jumbo data
Example of a jumbo event of 30 bytes in total, with 14 bytes of jumbo
data:
```
13 56 59 63 eb c1 4b 1a 96 d0 12 00 0e 00 00 00 |.VYc..K.........|
01 00 00 00 74 65 73 74 74 79 70 65 31 00 |....testtype1.|
```
In the following figure you can see each field annotated:
![Jumbo event](fig/event-jumbo.svg)
## Design considerations
The stream format has been designed to be very simple, so writing a
parser library would take no more than 2 days for a single developer.
The size of the events has been designed to be small, with 12 bytes per
event when no payload is used.
!!! Caution
The events are stored in disk following the endianness of the
machine where they are generated. So a stream generated with a
little endian machine would be different than on a big endian
machine. We assume the same endiannes is used to write the trace
at runtime and read it after, at the emulation process.
The events are designed to be easily identified when looking at the
raw stream in binary, as the MCV codes can be read as ASCII characters:
```
00000000 6f 76 6e 69 01 00 00 00 0f 4f 48 78 08 ba 2e 5c |ovni.....OHx...\|
00000010 b5 b0 00 00 00 00 00 00 ff ff ff ff 00 00 00 00 |................|
00000020 00 00 00 00 13 56 59 63 3c c2 2e 5c b5 b0 00 00 |.....VYc<..\....|
00000030 0e 00 00 00 01 00 00 00 74 65 73 74 74 79 70 65 |........testtype|
00000040 31 00 07 56 54 63 43 cc 2e 5c b5 b0 00 00 01 00 |1..VTcC..\......|
00000050 00 00 01 00 00 00 03 56 54 78 03 cd 2e 5c b5 b0 |.......VTx...\..|
00000060 00 00 01 00 00 00 03 56 54 70 2b 7d 37 5c b5 b0 |.......VTp+}7\..|
00000070 00 00 01 00 00 00 03 56 54 72 c3 4d 40 5c b5 b0 |.......VTr.M@\..|
00000080 00 00 01 00 00 00 03 56 54 65 03 36 49 5c b5 b0 |.......VTe.6I\..|
00000090 00 00 01 00 00 00 00 4f 48 65 f5 36 49 5c b5 b0 |.......OHe.6I\..|
000000a0 00 00 |..|
```
This allows a human to detect signs of corruption by visually inspecting
the streams.
## Limitations
The streams are designed to be read only forward, as they only contain
the size of each event in the header.
Currently, we only support using the threads as sources of events, using
one stream per thread. However, adding support for more streams from
multiple sources is planned for the future.