Questions tagged [infiniband]

InfiniBand is a high-speed switched fabric communications link technology used in high-performance computing and enterprise data centers.

InfiniBand is a switched fabric communications link used in high-performance computing and enterprise data centers. Its features include scalability, high throughput, low latency, quality of service and failover. The InfiniBand architecture specification defines a connection between processor nodes and high performance I/O nodes such as storage devices. Infiniband host bus adapters and network switches are commonly manufactured by both Mellanox and Intel.

178 questions
2
votes
1 answer

Mapping remote memory into the address space of a host using Inifiniband

I recently started to work with Infiniband cards, two Mellanox Technologies MT27700 Family [ConnectX-4] to be specific. Eventually, I want to extend an existing framework with interfaces based on the VPI Verbs API/RDMA CM API. About the research I…
Silicon1602
  • 1,151
  • 1
  • 7
  • 18
2
votes
1 answer

Mvapich2 (with IB) strange info after computing

I'm running example programs (hello-world & pi) with mvapich2 2.2. After computing I see info that looks like debug (in error thread). output thread Process 2 of 5 is on 37eff7e817ee Process 0 of 5 is on cb1479885879 Process 4 of 5 is on…
Ivister
  • 29
  • 3
2
votes
1 answer

MVAPICH2 RDMA-based communication without explicit PUT/GET use?

My cluster utilizes MVAPICH2 over Infiniband FDR and and I am considering the use of RDMA for my simulations. I am aware of the MPI_Put and MPI_Get calls for explicitly invoking RDMA operations, however I would like to know if this is the only way…
Tokth
  • 31
  • 4
2
votes
1 answer

What's the easiest way to use infiniband from .NET?

I've got two computers with Mellanox ConnectX VPI MT26428 infiniband cards running Windows 7 with drivers installed. What is the easiest way to start doing message passing (e.g. via MPI) from .NET?
J D
  • 48,105
  • 13
  • 171
  • 274
2
votes
2 answers

InfiniBand RDMA

I am trying to use the InfiniBand on an A8 machine on Azure. Actually, the ping-pong test works fine however, I can not run a simple RDMA based program. I can find the device by ibv_get_device_list(NULL) but I can not open it by the…
masoud
  • 31
  • 2
2
votes
0 answers

ibv_create_qp failure: is there a kernel parameter I can increase?

I'm using an acceleration library for my networking card to get kernel bypass while reading UDP packets. It works well, though when I subscribe beyond a certain number of sockets, the library calls into ibv_create_qp and fails with errno = ENOMEM.…
John S
  • 3,035
  • 2
  • 18
  • 29
2
votes
0 answers

How to develop with rsocket

I'm trying to use RDMA over infiniband as part of an application. I've discovered rsocket which looks like a useful API for RDMA. However I'm trying to use it but I have no idea what library I'm supposed to link with. I have rdma/rsocket.h included…
2
votes
1 answer

Setting up GPUDirect for infiniband

I try to setup GPUDirect to use infiniband verbs rdma calls directly on device memory without the need to use cudaMemcpy. I have 2 machines with nvidia k80 gpu cards each with driver version 367.27. CUDA8 is installed and Mellanox OFED 3.4 Also the…
kusterl
  • 29
  • 5
2
votes
1 answer

increase Memory Translation Table (MTT) for mellanox Connect-IB card

I have a fat node which has 2TB memory. With the new Connect-IB card, I want to increase the MTT, so I could register a large memory region. I found this post for HowTo Increase Memory Size used by Mellanox Adapters , but it didn't mention how to…
Zack
  • 1,205
  • 2
  • 14
  • 38
2
votes
1 answer

Infiniband vs. Gigabit ethernet, how do I control which is used by an MPI program

I have an MPI program that runs on a computer cluster that has both ethernet and Infiniband connectivity. When I compile with mpavich2's mpicc, it automatically links to the Infiniband libraries. Is there a way to control which network is used…
irritable_phd_syndrome
  • 4,631
  • 3
  • 32
  • 60
2
votes
1 answer

Not enough locked memory (rdma_setup_startup_ring: cannot create cq)

I have an issue with the quantity of available memory in page locked on CentOS 7. After allocating nodes with slurm, when I launch a job with MPI (mvapich), I encounter the following error: Fatal error in MPI_Init: Other MPI error, error…
jyvet
  • 2,021
  • 15
  • 22
2
votes
0 answers

How to bind my linux kernel module to InfiniBand HCA ports to send and receive messages via this module?

I want my kernel module get bind to an InfiniBand port and in turn register to the subnet administrator for future operations like query, send , receive. I have searched on the web but unfortunately there is very little info about it. At present,…
user3243499
  • 2,953
  • 6
  • 33
  • 75
2
votes
1 answer

What do I need to build to directly access the InfiniBand HCA ports and inject an IPoIB frame bits?

I would like to retrieve the IPoIB frame bits for all the IPoIB frames on the fabric no matter if they are destined (LID + QPN level) for my machine or not. Also, I should be able to re-inject the modified IPoIB frames directly to the InfiniBand HCA…
Utkal Sinha
  • 1,021
  • 1
  • 6
  • 24
2
votes
1 answer

what does mellanox interrupt mlx4-async@pci:0000 ... means?

I'm using an InfiniBand Mellanox card [ConnectX VPI PCIe 2.0 5GT/s - IB QDR / 10GigE] with OFED version 4-1.0.0 on an ubuntu 3.13.0 running on a x86_64 computer with 4 cores. Here is the result of ibstat on my computer CA 'mlx4_0' CA type:…
Fopa Léon Constantin
  • 11,863
  • 8
  • 48
  • 82
2
votes
1 answer

RDMA program randomly hangs

Anyone out there who has done RDMA programming using the RDMA_CM library? I'm having a hard time finding even simple examples to study. There's an rdma_client & rdma_server example in librdmacm, but it doesn't run in a loop (rping does loop, but…
Yale Zhang
  • 1,447
  • 12
  • 30