From: Zhu Yanjun <yanjun.zhu@linux.dev>
To: Michael Guralnik <michaelgur@nvidia.com>,
leonro@nvidia.com, jgg@nvidia.com
Cc: linux-rdma@vger.kernel.org, saeedm@nvidia.com, tariqt@nvidia.com
Subject: Re: [PATCH rdma-next 0/8] Introduce mlx5 Memory Scheme ODP
Date: Mon, 9 Sep 2024 10:10:43 +0800 [thread overview]
Message-ID: <faac55ee-6bfd-469f-a738-b0141b8d1b1e@linux.dev> (raw)
In-Reply-To: <d0f1e3c2-ac81-4126-bccd-615cd6ed3453@nvidia.com>
在 2024/9/8 14:18, Michael Guralnik 写道:
>
> On 06/09/2024 08:35, Zhu Yanjun wrote:
>>
>> As such, it seems that it can save memory via not pinning down the
>> underlying physical pages of the address space, and track the validity
>> of the mappings.
>>
>> What is the difference on the performance with/without ODP enabled? And
>> about memory usage, is there any test result about this?
>>
>> And ODP can be used mlx5 IB device? Or ODP can only be used in mlx5
>> RoCEv2 device?
>>
> The performance while using ODP is highly dependent on many factors that
> dictate how many page faults the kernel will have to deal with.
> Each page fault will introduce a latency hit.
>
> Both the examples in rdma_core (e.g ibv_rc_pingpong) and the perftest
> (e.g. ib_write_bw) support running with ODP to test this.
Thanks a lot. I have developed ODP for other RDMA devices. From my
tests, it seems that with ODP, the system memory is needed less than
without ODP.
From your descriptions, it seems that the latency of RDMA will increase
if I get you correctly.
If others (for example, bandwidth) remain unchanged, the tradeoff should
be between memory and latency.
Best Regards,
Zhu Yanjun
>
> ODP can be used in both IB and RoCE.
>
>
> Michael
>
>
>> Thanks,
>> Zhu Yanjun
>>
prev parent reply other threads:[~2024-09-09 2:10 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-04 15:30 [PATCH rdma-next 0/8] Introduce mlx5 Memory Scheme ODP Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 1/8] net/mlx5: Expand mkey page size to support 6 bits Michael Guralnik
2024-09-04 16:18 ` Jason Gunthorpe
2024-09-05 20:48 ` Michael Guralnik
2024-09-05 23:21 ` Jason Gunthorpe
2024-09-04 15:30 ` [PATCH rdma-next 2/8] net/mlx5: Expose HW bits for Memory scheme ODP Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 3/8] RDMA/mlx5: Add new ODP memory scheme eqe format Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 4/8] RDMA/mlx5: Enforce umem boundaries for explicit ODP page faults Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 5/8] RDMA/mlx5: Split ODP mkey search logic Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 6/8] RDMA/mlx5: Add handling for memory scheme page fault events Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 7/8] RDMA/mlx5: Add implicit MR handling to ODP memory scheme Michael Guralnik
2024-09-04 15:30 ` [PATCH rdma-next 8/8] net/mlx5: Handle memory scheme ODP capabilities Michael Guralnik
2024-09-06 5:35 ` [PATCH rdma-next 0/8] Introduce mlx5 Memory Scheme ODP Zhu Yanjun
2024-09-08 6:18 ` Michael Guralnik
2024-09-09 2:10 ` Zhu Yanjun [this message]
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=faac55ee-6bfd-469f-a738-b0141b8d1b1e@linux.dev \
--to=yanjun.zhu@linux.dev \
--cc=jgg@nvidia.com \
--cc=leonro@nvidia.com \
--cc=linux-rdma@vger.kernel.org \
--cc=michaelgur@nvidia.com \
--cc=saeedm@nvidia.com \
--cc=tariqt@nvidia.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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.