Improving Hwgraph to support large rendering #97
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ErwanAliasr1
force-pushed
the
hwgraph-updates
branch
3 times, most recently
from
5786e11 to
bc67c4b
Compare
anisse
reviewed
Sep 1, 2025
anisse
reviewed
Sep 1, 2025
anisse
reviewed
Sep 1, 2025
anisse
reviewed
Sep 1, 2025
|
Again for the --same-rack, I'm not sure I understand why the servers should be in the same rack for this graph? Why does it matter? At least for --same-chassis, we assume they share the PSUs. What do they share in --same-rack? Why can't it be any arbitrary --group-of-servers ? |
When plotting large graphs with crossing or nearly parallel lines, using the "o" marker makes a too large marker on the line. As a consequence, the graph is hard to read and some part of the lines are nearly unreadable. This commit is using the smallest marker which is large enough to be seen, increasing th readbility of the graphs. Signed-off-by: Erwan Velu <[email protected]>
The lower-case 's' makes no sense here. Signed-off-by: Erwan Velu <[email protected]>
ErwanAliasr1
force-pushed
the
hwgraph-updates
branch
2 times, most recently
from
df80e9b to
5c2785e
Compare
The current directory structure was like :
- <host1>
- <host2>
- <host.n>
- individual
- scaling
This was a bit confusing to read as per the names and structure.
This commit is offering a new scheme like:
- environment / { by_host | by_chassis} / {host1, host2, host.n} / {metric1, metric2, ...} / { graph1, graph2, ...}
- max_versus / { metric1, metric2, ...} / {job1, job2, ...} / { graph1, graph2, ...}
- smp_scaling / {metric1, metric2, ...} / {graph1, graph2, ...}
To match this new semantic, --no-individual argument is renamed --no-versus
Signed-off-by: Erwan Velu <[email protected]>
Some benchmarks are operated with external scenario like: - shutdown a PSU during a period of time - remove liquid cooling during a portion of a run - remove a fan .... While rendering the graphs, it was not yet possible to indicate specific period of the benchmark where some external events occured. This commit adds a --events option to hwgraph to add to the graph some visual indications that events occured at some period of the benchmark. Each event is declared like <event_name>:<start_time>:<duration>, a example looks like the following : uv run hwgraph graph --traces mydir/results.json:my_server_name:BMC.Server --title 'Full CPU load 10 minutes - no coolant for 90 secs' --outdir mydir/graphs --events no-coolant:0:90 In this example, a background nearly transparent box will be added for the first 90 seconds (start_time=0s, duration=90s) of the benchmark. A new entry will be added to the legend : "Event no-coolant" As per the --traces option, multiple events can be defined, each new event will be colored differently to ensure an easy reading. Signed-off-by: Erwan Velu <[email protected]>
Signed-off-by: Erwan Velu <[email protected]>
When too many boxes are rendered at once, the horizontal print is unreadble. If more than 10 boxes are rendered, let's switch in vertical mode for the associated label. Signed-off-by: Erwan Velu <[email protected]>
ErwanAliasr1
force-pushed
the
hwgraph-updates
branch
from
5c2785e to
b573fd1
Compare
When performing full rack benchmarking, it could be interesting to graph some specific graphs for studying the following metrics during the full run: - the sum of the servers power consumption - the sum of the cpus power consumption - the sum of difference between server's power and cpu power consumptions This patch also prints the percentage ratio of the cpu power and the other components versus the server's power consumption. It could be interesting to see how much of a server power is consumed by the CPUs. This is about the same idea as the --same-chassis but at the scale of a group of servers. This feature, will be useful to graph results of a given rack. Signed-off-by: Erwan Velu <[email protected]>
The max_versus mode is useful to compare the same run over several machines or the same machine in various conditions. If a single trace is provided, the produced graphs are not very interesting to read. Let's disable them and print a message to explain the reason of this automatic feature disabling. Signed-off-by: Erwan Velu <[email protected]>
The current version of hwgraph crashes if the trace are missing PDU or IPC metrics in the trace file. This patch ignores and report when a metric is missing instead of crashing with KeyError exceptions. This improves hwgraph's retrocompatiblity with older traces. Signed-off-by: Erwan Velu <[email protected]>
The current code was plotting each iteration of the benchmark which generated tons of nearly useless graphs. This commit keeps the maximum values and drops intermediates ones. Signed-off-by: Erwan Velu <[email protected]>
time_to_next_sync() returned the number of seconds before the next meeting point but also the "next_sync" absolute time. As the next_sync is not used at all, let's remove is useless return value and simplify the return code function. Signed-off-by: Erwan Velu <[email protected]>
ErwanAliasr1
force-pushed
the
hwgraph-updates
branch
from
d80648f to
865b1d8
Compare
|
repushed sorry, forgot to fix typos you reported :/ |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request is about bug fixing hwgraph and add a few features to ease the rendering of large datasets.