BGP: Improve tx performance during feed/flush

The prefix hash table in BGP used the same hash function as the rtable.
When a batch of routes are exported during feed/flush to the BGP, they
all have similar hash values, so they are all crowded in a few slots in
the BGP prefix table (which is much smaller - around the size of the
batch - and uses higher bits from hash values), making it much slower due
to excessive collisions. Use a different hash function to avoid this.

Also, increase the batch size to fill 4k BGP packets and increase minimum
BGP bucket and prefix hash sizes to avoid back and forth resizing during
flushes.

This leads to order of magnitude faster flushes (on my test data).
This commit is contained in:
Ondrej Zajicek 2022-05-15 15:05:13 +02:00
parent 207ac48533
commit ba2a076001
2 changed files with 5 additions and 4 deletions

View file

@ -2380,7 +2380,7 @@ static void
rt_prune_table(rtable *tab)
{
struct fib_iterator *fit = &tab->prune_fit;
int limit = 512;
int limit = 2000;
struct channel *c;
node *n, *x;

View file

@ -1458,7 +1458,7 @@ bgp_finish_attrs(struct bgp_parse_state *s, rta *a)
#define RBH_FN(a,h) h
#define RBH_REHASH bgp_rbh_rehash
#define RBH_PARAMS /8, *2, 2, 2, 8, 20
#define RBH_PARAMS /8, *2, 2, 2, 12, 20
HASH_DEFINE_REHASH_FN(RBH, struct bgp_bucket)
@ -1600,7 +1600,7 @@ bgp_withdraw_bucket(struct bgp_channel *c, struct bgp_bucket *b)
#define PXH_FN(n,i,h) h
#define PXH_REHASH bgp_pxh_rehash
#define PXH_PARAMS /8, *2, 2, 2, 8, 24
#define PXH_PARAMS /8, *2, 2, 2, 12, 24
HASH_DEFINE_REHASH_FN(PXH, struct bgp_prefix)
@ -1626,7 +1626,8 @@ bgp_free_prefix_table(struct bgp_channel *c)
static struct bgp_prefix *
bgp_get_prefix(struct bgp_channel *c, net_addr *net, u32 path_id)
{
u32 hash = net_hash(net) ^ u32_hash(path_id);
/* We must use a different hash function than the rtable */
u32 hash = u32_hash(net_hash(net) ^ u32_hash(path_id));
struct bgp_prefix *px = HASH_FIND(c->prefix_hash, PXH, net, path_id, hash);
if (px)