Menu
Page Index

Blame

be9f18 Lorphos 2026-03-07 23:06:58
initial version, WIP
1 
# Clustering with RDMA
2
3
## Clustering with Oculink and PCIe 3.0 Infiniband cards
4
5
The more recent PCIe 4.0 cards are quite a bit more expensive than the older cards. The PCIe 3.0 x4 connection limits the cards to speeds of around 26GByte/s. Not too shabby.
6
7
Here's some hardware used for a setup with cheap used Mellanox cards:
8
9
* 2x Strix Halo with a spare M.2 slot (tested using Bosgame M5)
10
* 1x ATX PC PSU (any will do, needs just 20 Watts)
11
* 2x Mellanox ConnectX-3 CX354A PCIe 3.0 x8 infiniband cards, used, 23€ each [example link](https://www.ebay.de/itm/177760210929?_skw=cx354a&epid=7043214331&itmmeta=01KK55RMHKERWWE085D2FZ4C5G&hash=item2963558ff1:g:u1MAAeSw7MRpYSrx)
12
* 1x DAC cable Mellanox 56G QSFP+ FDR InfiniBand DAC Copper Twinax Passiv 0.5m MC2207130-00A, used, 18€ [example link](https://www.ebay.de/itm/126922287689)
13
* 1x ATX PSU 24pin splitter cable [example link](https://a.aliexpress.com/_Ezm7My8) ($6 with coins)
14
* 2x Oculink M.2 adapter, capble, PCIe 4.0 x16 slot [example link](https://a.aliexpress.com/_Ez9CgPK) (~$25 each with coins and coupons)
15
16
Total cost: 46€+18€+49€ =113€ Not bad!
17
18
What else is needed:
19
20
* a little 3d printed custom case for the two network cards
21
* 2x 3d printed covers for the SSD compartment with a hole for the Oculink cable. Or you drill a whole in the original metal covers.
22
* a little fan to keep the Mellanox cards cool inside the case (they use up to 10W each)
23
24
### Quick howto:
25
26
1. Connect Oculink M.2 adapters to the empty M.2 NVMe slots (1 per PC).
27
2. Plug Oculink cables into M.2 adapters and into PCIe 4.0 x16 slot adapters.
28
3. Plug 24pin PSU split cable into both PCIe 4.0 x16 slot adapters and into PSU.
29
4. Plug the two Mellanox cards into the PCIe slots
30
5. Connect the two Mellanox cards with the DAC cable
31
6. Using the switch on the PCIe 4.0 x16 slot adapter, turn on the PSU.
32
7. Finally, turn on the PCs.
33
34
Check if you can see the Mellanox cards in `lspci`:
35
36
```$ lspci
37
…
38
c3:00.0 Network controller: Mellanox Technologies MT27500 Family [ConnectX-3]
39
…
40
```
41
Make sure the NIC is connected via PCIe 3.0 x4:
42
```$ sudo lspci -vv -s c3:00.0 |grep -E "LnkCap:|LnkSta:"
43
LnkCap: Port #8, Speed 8GT/s, Width x8, ASPM L0s, Exit Latency L0s unlimited
44
LnkSta: Speed 8GT/s, Width x4 (downgraded)
45
```
46
Install needed packages on both PCs running Fedora 43:
47
```$ sudo dnf install rdma-core libibverbs-utils mstflint infiniband-diags perftest
48
$ ibv_devinfo
49
```
50
look for „Link Layer“, it should show Infiniband
51
on PC1 we start opensm, the Infiniband subnet manager and administration:
52
```$ sudo dnf install opensm
53
$ sudo systemctl enable --now opensm
54
$ sudo restorecon -v /var/log/opensm.log
55
56
$ ibstat
57
```
58
now shows „State: Active“ on both PCs
59
60
PC1:
61
```$ ip a|grep -B 1 infini
62
4: ibp195s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 2044 qdisc fq_codel state UP group default qlen 1000
63
link/infiniband 80:00:02:08:fe:80:00:00:00:00:00:00:ec:0d:9a:03:00:xx:xx:xx brd 00:ff:ff:ff:ff:12:40:1b:ff:ff:00:00:00:00:00:00:ff:ff:ff:ff
64
5: ibp195s0d1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 4092 qdisc fq_codel state DOWN group default qlen 1000
65
link/infiniband 80:00:02:09:fe:80:00:00:00:00:00:00:ec:0d:9a:03:00:xx:xx:xx brd 00:ff:ff:ff:ff:12:40:1b:ff:ff:00:00:00:00:00:00:ff:ff:ff:ff
66
```
67
PC2:
68
```$ ip a|grep -B 1 infini
69
3: ibp195s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 2044 qdisc fq_codel state UP group default qlen 1000
70
link/infiniband 80:00:02:08:fe:80:00:00:00:00:00:00:ec:0d:9a:03:00:yy:yy:yy brd 00:ff:ff:ff:ff:12:40:1b:ff:ff:00:00:00:00:00:00:ff:ff:ff:ff
71
4: ibp195s0d1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 4092 qdisc fq_codel state DOWN group default qlen 1000
72
link/infiniband 80:00:02:09:fe:80:00:00:00:00:00:00:ec:0d:9a:03:00:yy:yy:yy brd 00:ff:ff:ff:ff:12:40:1b:ff:ff:00:00:00:00:00:00:ff:ff:ff:ff
73
```
74
So the interface name is **ibp195s0** on both PCs.
75
76
configure IPv4 on PC1:
77
```$ sudo nmcli conn add type infiniband con-name ib-conn ifname ibp195s0 transport-mode datagram ipv4.method manual ipv4.addresses 192.168.100.1/24
78
Verbindung »ib-conn« (e6655fba-ebd6-4ee5-a31b-9c25faacfe37) erfolgreich hinzugefügt.
79
```
80
configure IPv4 on PC2:
81
```$ sudo nmcli conn add type infiniband con-name ib-conn ifname ibp195s0 transport-mode datagram ipv4.method manual ipv4.addresses 192.168.100.2/24
82
$ sudo nmcli conn up ib-conn
83
$ sudo nmcli conn show
84
```
85
PC1: (I also have a connection via Thunderbolt)
86
```$ sudo nmcli conn up ib-conn
87
$ sudo nmcli conn show
88
NAME UUID TYPE DEVICE
89
Kabelgebundene Verbindung 1 1a44c330-8d06-34d6-9773-df0a34882a4b ethernet eno1
90
ib-conn e6655fba-ebd6-4ee5-a31b-9c25faacfe37 infiniband ibp195s0
91
thunderbolt0 7beaa789-b367-4810-ba22-3e946edab0fd ethernet thunderbolt0
92
```
93
PC2:
94
```$ sudo nmcli conn show
95
NAME UUID TYPE DEVICE
96
Kabelgebundene Verbindung 1 dea9361f-0f51-3acf-9b85-04a35c116b67 ethernet eno1
97
ib-conn 5eaa86fe-99e7-48c9-b460-740d31adc936 infiniband ibp195s0
98
thunderbolt0 bd7e1a3c-f05d-3a43-bfc0-880fb874dba4 ethernet thunderbolt0
99
```
100
Check with „ip a“ if the infiniband interfaces are up. If not, check on PC1 if opensm is giving errors?
101
OK, if the connection is up, we can check the bandwidth:
102
on PC1:
103
```$ ib_write_bw
104
```
105
on PC2:
106
```$ ib_write_bw 192.168.100.1
107
#bytes #iterations BW peak[MiB/sec] BW average[MiB/sec] MsgRate[Mpps]
108
65536 5000 3293.63 3293.56 0.052697
109
```
110
and we can check the latency:
111
on PC1:
112
```$ ib_write_lat
113
```
114
on PC2:
115
```$ ib_write_lat 192.168.100.1
116
#bytes #iterations t_min[usec] t_max[usec] t_typical[usec] t_avg[usec] t_stdev[usec] 99% percentile[usec] 99.9% percentile[usec]
117
2 1000 1.10 2.05 1.11 1.12 0.00 1.19 2.05
118
```
119
So around 1.12µs which is an expected value.
120
121
Next, follow the [AMD Strix Halo RDMA Cluster Setup Guide](https://github.com/kyuz0/amd-strix-halo-vllm-toolboxes/blob/main/rdma_cluster/setup_guide.md)
122