DPDK patches and discussions
 help / color / mirror / Atom feed
From: Georgios Katsikas <katsikas.gp@gmail.com>
To: wisamm@mellanox.com
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH] app/flow-perf: configurable rule batches
Date: Tue, 3 Nov 2020 13:26:03 +0200
Message-ID: <CAN9HtFCpYtwqa+nySWvUKmQP8MXU5skASpMd7gE2ytm4UxyCkA@mail.gmail.com> (raw)
In-Reply-To: <20201011100326.9074-1-katsikas.gp@gmail.com>

Hi,

Any news on this patch?
Is there anything else I could do?

Thanks,
Georgios

On Sun, Oct 11, 2020 at 1:03 PM Georgios Katsikas <katsikas.gp@gmail.com>
wrote:

> Currently, flow-perf measures the performance of
> rule installation/deletion operations by breaking
> down the entire number of operations into windows
> of fixed size (i.e., 100000 operations per window).
> Then, flow-perf measures the total time per window
> and computes an average time across all windows.
>
> This commit allows flow-perf users to configure
> the number of rules per window instead of using
> a fixed pre-compiled value. To do so, users must
> pass --rules-batch=N, where N is the number of
> rules per window (or batch).
> For consistency reasons, flow_count variable is
> now renamed to rules_count. This variable is the
> total number of rules to be installed/deleted.
>
> For example, if a user wants to measure how much
> time it takes to install 1M rules in a certain NIC,
> he/she can input:
> --rules-count=1000000
> This way flow-perf will break down 1M flow rules into
> 10 batches of 100k flow rules each (this is the default
> batch size) and compute an average across the 10
> measurements.
> Now, if the user modifies the number of rules per
> batch as follows:
> --rules-count=1000000 --rules-batch=500000
> then flow-perf will break down 1M flow rules into
> 2 batches of 500k flow rules each and compute the
> average across the 2 measurements.
>
> Finally, this commit also adds default variables
> to the usage function instead of hardcoded values.
>
> Signed-off-by: Georgios Katsikas <katsikas.gp@gmail.com>
> ---
>  app/test-flow-perf/main.c      | 86 ++++++++++++++++++++--------------
>  doc/guides/tools/flow-perf.rst | 42 ++++++++++++-----
>  2 files changed, 79 insertions(+), 49 deletions(-)
>
> diff --git a/app/test-flow-perf/main.c b/app/test-flow-perf/main.c
> index c420da6a5..4cdab2c93 100644
> --- a/app/test-flow-perf/main.c
> +++ b/app/test-flow-perf/main.c
> @@ -40,7 +40,8 @@
>
>  #define MAX_ITERATIONS             100
>  #define DEFAULT_RULES_COUNT    4000000
> -#define DEFAULT_ITERATION       100000
> +#define DEFAULT_RULES_BATCH     100000
> +#define DEFAULT_GROUP                0
>
>  struct rte_flow *flow;
>  static uint8_t flow_group;
> @@ -62,8 +63,8 @@ static bool enable_fwd;
>
>  static struct rte_mempool *mbuf_mp;
>  static uint32_t nb_lcores;
> -static uint32_t flows_count;
> -static uint32_t iterations_number;
> +static uint32_t rules_count;
> +static uint32_t rules_batch;
>  static uint32_t hairpin_queues_num; /* total hairpin q number - default:
> 0 */
>  static uint32_t nb_lcores;
>
> @@ -98,8 +99,10 @@ usage(char *progname)
>  {
>         printf("\nusage: %s\n", progname);
>         printf("\nControl configurations:\n");
> -       printf("  --flows-count=N: to set the number of needed"
> -               " flows to insert, default is 4,000,000\n");
> +       printf("  --rules-count=N: to set the number of needed"
> +               " rules to insert, default is %d\n", DEFAULT_RULES_COUNT);
> +       printf("  --rules-batch=N: set number of batched rules,"
> +               " default is %d\n", DEFAULT_RULES_BATCH);
>         printf("  --dump-iterations: To print rates for each"
>                 " iteration\n");
>         printf("  --deletion-rate: Enable deletion rate"
> @@ -114,7 +117,7 @@ usage(char *progname)
>         printf("  --egress: set egress attribute in flows\n");
>         printf("  --transfer: set transfer attribute in flows\n");
>         printf("  --group=N: set group for all flows,"
> -               " default is 0\n");
> +               " default is %d\n", DEFAULT_GROUP);
>
>         printf("To set flow items:\n");
>         printf("  --ether: add ether layer in flow items\n");
> @@ -527,7 +530,8 @@ args_parse(int argc, char **argv)
>         static const struct option lgopts[] = {
>                 /* Control */
>                 { "help",                       0, 0, 0 },
> -               { "flows-count",                1, 0, 0 },
> +               { "rules-count",                1, 0, 0 },
> +               { "rules-batch",                1, 0, 0 },
>                 { "dump-iterations",            0, 0, 0 },
>                 { "deletion-rate",              0, 0, 0 },
>                 { "dump-socket-mem",            0, 0, 0 },
> @@ -705,16 +709,26 @@ args_parse(int argc, char **argv)
>                         }
>                         /* Control */
>                         if (strcmp(lgopts[opt_idx].name,
> -                                       "flows-count") == 0) {
> +                                       "rules-batch") == 0) {
>                                 n = atoi(optarg);
> -                               if (n > (int) iterations_number)
> -                                       flows_count = n;
> +                               if (n >= DEFAULT_RULES_BATCH)
> +                                       rules_batch = n;
>                                 else {
> -                                       printf("\n\nflows_count should be
> > %d\n",
> -                                               iterations_number);
> +                                       printf("\n\nrules_batch should be
> >= %d\n",
> +                                               DEFAULT_RULES_BATCH);
>                                         rte_exit(EXIT_SUCCESS, " ");
>                                 }
>                         }
> +                       if (strcmp(lgopts[opt_idx].name,
> +                                       "rules-count") == 0) {
> +                               n = atoi(optarg);
> +                               if (n >= (int) rules_batch)
> +                                       rules_count = n;
> +                               else {
> +                                       printf("\n\nrules_count should be
> >= %d\n",
> +                                               rules_batch);
> +                               }
> +                       }
>                         if (strcmp(lgopts[opt_idx].name,
>                                         "dump-iterations") == 0)
>                                 dump_iterations = true;
> @@ -826,13 +840,13 @@ destroy_flows(int port_id, struct rte_flow
> **flow_list)
>         for (i = 0; i < MAX_ITERATIONS; i++)
>                 cpu_time_per_iter[i] = -1;
>
> -       if (iterations_number > flows_count)
> -               iterations_number = flows_count;
> +       if (rules_batch > rules_count)
> +               rules_batch = rules_count;
>
>         /* Deletion Rate */
>         printf("Flows Deletion on port = %d\n", port_id);
>         start_iter = clock();
> -       for (i = 0; i < flows_count; i++) {
> +       for (i = 0; i < rules_count; i++) {
>                 if (flow_list[i] == 0)
>                         break;
>
> @@ -842,11 +856,11 @@ destroy_flows(int port_id, struct rte_flow
> **flow_list)
>                         rte_exit(EXIT_FAILURE, "Error in deleting flow");
>                 }
>
> -               if (i && !((i + 1) % iterations_number)) {
> +               if (i && !((i + 1) % rules_batch)) {
>                         /* Save the deletion rate of each iter */
>                         end_iter = clock();
>                         delta = (double) (end_iter - start_iter);
> -                       iter_id = ((i + 1) / iterations_number) - 1;
> +                       iter_id = ((i + 1) / rules_batch) - 1;
>                         cpu_time_per_iter[iter_id] =
>                                 delta / CLOCKS_PER_SEC;
>                         cpu_time_used += cpu_time_per_iter[iter_id];
> @@ -859,21 +873,21 @@ destroy_flows(int port_id, struct rte_flow
> **flow_list)
>                 for (i = 0; i < MAX_ITERATIONS; i++) {
>                         if (cpu_time_per_iter[i] == -1)
>                                 continue;
> -                       delta = (double)(iterations_number /
> +                       delta = (double)(rules_batch /
>                                 cpu_time_per_iter[i]);
>                         flows_rate = delta / 1000;
>                         printf(":: Iteration #%d: %d flows "
>                                 "in %f sec[ Rate = %f K/Sec ]\n",
> -                               i, iterations_number,
> +                               i, rules_batch,
>                                 cpu_time_per_iter[i], flows_rate);
>                 }
>
>         /* Deletion rate for all flows */
> -       flows_rate = ((double) (flows_count / cpu_time_used) / 1000);
> +       flows_rate = ((double) (rules_count / cpu_time_used) / 1000);
>         printf("\n:: Total flow deletion rate -> %f K/Sec\n",
>                 flows_rate);
>         printf(":: The time for deleting %d in flows %f seconds\n",
> -               flows_count, cpu_time_used);
> +               rules_count, cpu_time_used);
>  }
>
>  static inline void
> @@ -902,13 +916,13 @@ flows_handler(void)
>         for (i = 0; i < MAX_ITERATIONS; i++)
>                 cpu_time_per_iter[i] = -1;
>
> -       if (iterations_number > flows_count)
> -               iterations_number = flows_count;
> +       if (rules_batch > rules_count)
> +               rules_batch = rules_count;
>
> -       printf(":: Flows Count per port: %d\n", flows_count);
> +       printf(":: Flows Count per port: %d\n", rules_count);
>
>         flow_list = rte_zmalloc("flow_list",
> -               (sizeof(struct rte_flow *) * flows_count) + 1, 0);
> +               (sizeof(struct rte_flow *) * rules_count) + 1, 0);
>         if (flow_list == NULL)
>                 rte_exit(EXIT_FAILURE, "No Memory available!");
>
> @@ -941,7 +955,7 @@ flows_handler(void)
>                 /* Insertion Rate */
>                 printf("Flows insertion on port = %d\n", port_id);
>                 start_iter = clock();
> -               for (i = 0; i < flows_count; i++) {
> +               for (i = 0; i < rules_count; i++) {
>                         flow = generate_flow(port_id, flow_group,
>                                 flow_attrs, flow_items, flow_actions,
>                                 JUMP_ACTION_TABLE, i,
> @@ -950,7 +964,7 @@ flows_handler(void)
>                                 &error);
>
>                         if (force_quit)
> -                               i = flows_count;
> +                               i = rules_count;
>
>                         if (!flow) {
>                                 print_flow_error(error);
> @@ -959,11 +973,11 @@ flows_handler(void)
>
>                         flow_list[flow_index++] = flow;
>
> -                       if (i && !((i + 1) % iterations_number)) {
> +                       if (i && !((i + 1) % rules_batch)) {
>                                 /* Save the insertion rate of each iter */
>                                 end_iter = clock();
>                                 delta = (double) (end_iter - start_iter);
> -                               iter_id = ((i + 1) / iterations_number) -
> 1;
> +                               iter_id = ((i + 1) / rules_batch) - 1;
>                                 cpu_time_per_iter[iter_id] =
>                                         delta / CLOCKS_PER_SEC;
>                                 cpu_time_used +=
> cpu_time_per_iter[iter_id];
> @@ -976,21 +990,21 @@ flows_handler(void)
>                         for (i = 0; i < MAX_ITERATIONS; i++) {
>                                 if (cpu_time_per_iter[i] == -1)
>                                         continue;
> -                               delta = (double)(iterations_number /
> +                               delta = (double)(rules_batch /
>                                         cpu_time_per_iter[i]);
>                                 flows_rate = delta / 1000;
>                                 printf(":: Iteration #%d: %d flows "
>                                         "in %f sec[ Rate = %f K/Sec ]\n",
> -                                       i, iterations_number,
> +                                       i, rules_batch,
>                                         cpu_time_per_iter[i], flows_rate);
>                         }
>
>                 /* Insertion rate for all flows */
> -               flows_rate = ((double) (flows_count / cpu_time_used) /
> 1000);
> +               flows_rate = ((double) (rules_count / cpu_time_used) /
> 1000);
>                 printf("\n:: Total flow insertion rate -> %f K/Sec\n",
>                                                 flows_rate);
>                 printf(":: The time for creating %d in flows %f seconds\n",
> -                                               flows_count,
> cpu_time_used);
> +                                               rules_count,
> cpu_time_used);
>
>                 if (delete_flag)
>                         destroy_flows(port_id, flow_list);
> @@ -1415,11 +1429,11 @@ main(int argc, char **argv)
>
>         force_quit = false;
>         dump_iterations = false;
> -       flows_count = DEFAULT_RULES_COUNT;
> -       iterations_number = DEFAULT_ITERATION;
> +       rules_count = DEFAULT_RULES_COUNT;
> +       rules_batch = DEFAULT_RULES_BATCH;
>         delete_flag = false;
>         dump_socket_mem_flag = false;
> -       flow_group = 0;
> +       flow_group = DEFAULT_GROUP;
>
>         signal(SIGINT, signal_handler);
>         signal(SIGTERM, signal_handler);
> diff --git a/doc/guides/tools/flow-perf.rst
> b/doc/guides/tools/flow-perf.rst
> index 7e5dc0c54..018358ac1 100644
> --- a/doc/guides/tools/flow-perf.rst
> +++ b/doc/guides/tools/flow-perf.rst
> @@ -5,19 +5,25 @@ Flow Performance Tool
>  =====================
>
>  Application for rte_flow performance testing.
> -The application provide the ability to test insertion rate of specific
> -rte_flow rule, by stressing it to the NIC, and calculate the insertion
> -rate.
> +The application provides the ability to test insertion rate of specific
> +rte_flow rule, by stressing it to the NIC, and calculates the insertion
> +and deletion rates.
>
> -The application offers some options in the command line, to configure
> -which rule to apply.
> +The application allows to configure which rule to apply through several
> +options of the command line.
>
>  After that the application will start producing rules with same pattern
>  but increasing the outer IP source address by 1 each time, thus it will
>  give different flow each time, and all other items will have open masks.
>
> -The application also provide the ability to measure rte flow deletion
> rate,
> -in addition to memory consumption before and after the flows creation.
> +To assess the rule insertion rate, the flow performance tool breaks
> +down the entire number of flow rule operations into windows of fixed size
> +(defaults to 100000 flow rule operations per window, but can be
> configured).
> +Then, the flow performance tool measures the total time per window and
> +computes an average time across all windows.
> +
> +The application also provides the ability to measure rte flow deletion
> rate,
> +in addition to memory consumption before and after the flow rules'
> creation.
>
>  The app supports single and multi core performance measurements.
>
> @@ -59,21 +65,31 @@ with a ``--`` separator:
>
>  .. code-block:: console
>
> -       sudo ./dpdk-test-flow_perf -n 4 -w 08:00.0 -- --ingress --ether
> --ipv4 --queue --flows-count=1000000
> +       sudo ./dpdk-test-flow_perf -n 4 -w 08:00.0 -- --ingress --ether
> --ipv4 --queue --rules-count=1000000
>
>  The command line options are:
>
>  *      ``--help``
>         Display a help message and quit.
>
> -*      ``--flows-count=N``
> -       Set the number of needed flows to insert,
> -       where 1 <= N <= "number of flows".
> +*      ``--rules-count=N``
> +       Set the total number of flow rules to insert,
> +       where 1 <= N <= "number of flow rules".
>         The default value is 4,000,000.
>
> +*      ``--rules-batch=N``
> +       Set the number of flow rules to insert per iteration window,
> +       where 1 <= N <= "number of flow rules per iteration window".
> +       The default value is 100,000 flow rules per iteration window.
> +       For a total of --rules-count=1000000 flow rules to be inserted
> +       and an iteration window size of --rules-batch=100000 flow rules,
> +       the application will measure the insertion rate 10 times
> +       (i.e., once every 100000 flow rules) and then report an average
> +       insertion rate across the 10 measurements.
> +
>  *      ``--dump-iterations``
> -       Print rates for each iteration of flows.
> -       Default iteration is 1,00,000.
> +       Print rates for each iteration window.
> +       Default iteration window equals to the rules-batch size (i.e.,
> 100,000).
>
>  *      ``--deletion-rate``
>         Enable deletion rate calculations.
> --
> 2.17.1
>
>

  reply	other threads:[~2020-11-03 11:26 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-11 10:03 Georgios Katsikas
2020-11-03 11:26 ` Georgios Katsikas [this message]
2020-11-04  8:04   ` Wisam Monther
2020-11-04 11:11     ` Georgios Katsikas
2020-11-04 11:25       ` Wisam Monther
2020-11-04 20:46     ` Thomas Monjalon
  -- strict thread matches above, loose matches on Subject: below --
2020-09-24  9:10 Georgios Katsikas
2020-09-24 12:01 ` Wisam Monther
2020-10-05 17:16   ` Georgios Katsikas
2020-10-05 22:25     ` Thomas Monjalon
2020-10-11 10:33       ` Georgios Katsikas

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAN9HtFCpYtwqa+nySWvUKmQP8MXU5skASpMd7gE2ytm4UxyCkA@mail.gmail.com \
    --to=katsikas.gp@gmail.com \
    --cc=dev@dpdk.org \
    --cc=wisamm@mellanox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ https://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git