
Sign up to save your podcasts
Or


In this episode of the backend engineering show I go through the main job of the network interface controller (NIC for short) and how the datacenter is pushing it to the limit by allowing it to do more TCP/IP processing, creating what is being popularized as smartNIC.
0:00 Intro
1:20 What is a NIC?
3:40 NIC job
8:00 When does the OS get involved
12:40 Promiscuous mode
14:00 SmartNIC
18:30 Disadvantages
Resources
https://developer.nvidia.com/networking/ethernet-adapters
https://www.theregister.com/2022/08/11/smartnics_network_market/
https://arxiv.org/abs/1803.09615
Fundamentals of Networking for Effective Backends udemy course (link redirects to udemy with coupon)
https://network.husseinnasser.com
In this episode of the backend engineering show I go through and discuss the Homa Protocol paper which attempts to replace TCP as a protocol in the data centers. I learned a lot from this paper, I have my criticisms of certain aspects, timestamps for topics discussed below.
It appears there is a path to replace TCP in the datacenter and professor John tries to explain this path.
Referenced materials mentioned in the episode
Overview paper
https://web.stanford.edu/~ouster/cgi-bin/papers/replaceTcp.pdf
Homa 2018 paper (Details)
https://people.csail.mit.edu/alizadeh/papers/homa-sigcomm18.pdf
NIC Offloading in Linux
https://en.wikipedia.org/wiki/TCP_offload_engine#Support_in_Linux
Curl disabling Nigel Algo
https://github.com/curl/curl/commit/4732ca5724072f132876f520c8f02c7c5b654d9
0:00 Intro
3:00 The nature of networking data center
5:30 TCP Segments
7:30 There is no “Request” in TCP
12:00 What so unique about Data centers?
14:00 Message Throughput vs Data throughput
18:25 Congestion Control
22:38 Homa’s Congestion Control
25:00 Server Core Load Balancing
28:30 NIC offloading
30:00 Everything Wrong about TCP
37:00 Why not QUIC?
40:00 Limitation of Streaming
44:10 Load Balancing Stream Reading
47:15 Can we treat Segments as Messages?
51:00 Dispatching Messages is Easier
53:00 Connection Orientation
1:00:00 Sender Driven Congestion Control
1:03:00 In Order Packet Delivery
1:07:00 DCTCP
1:08:30 Homa is Message Based
1:11:00 Home is Connection Less
1:12:00 Receiver Driven Congestion Control
1:15:19 Out of Order Packets
1:16:20 Homa API is not Compatible with TCP
1:17:40 Will Homa come to HTTP?
1:18:45 Conclusion
ByteDance, the company behind TikTok is proposing few patches to the linux kernel to make kernel reboots via kexec go from 500ms down to 15 ms saving huge time in updating kernel on thousands of machines. Let us discuss this
0:00 Intro
1:30 Linux Kernel Reboot Options
2:30 how kexec works
4:00 The optimization
5:00 Going through the patch
6:00 Updating Servers at TikTok scale
9:00 Summary
https://www.phoronix.com/news/Bytedance-Faster-Kexec-Reboot
https://lore.kernel.org/lkml/[email protected]/
We discuss the CVE-2022-2274 OpenSSL Vulnerability.
The OpenSSL 3.0.4 release introduced a serious bug in the RSA
implementation for X86_64 CPUs supporting the AVX512IFMA instructions.
This issue makes the RSA implementation with 2048 bit private keys
incorrect on such machines and memory corruption will happen during
the computation. As a consequence of the memory corruption an attacker
may be able to trigger a remote code execution on the machine performing
the computation.
0:00 Intro
1:00 CVE-2022-2274
3:00 AVX512IFMA CISC
5:00 How the bug works
7:10 How can it be triggered
Resources
https://www.openssl.org/news/secadv/20220705.txt
https://github.com/openssl/openssl/issues/18625
https://guidovranken.com/2022/06/27/notes-on-openssl-remote-memory-corruption/
https://eprint.iacr.org/2018/335
https://github.com/openssl/openssl/commit/4d8a88c134df634ba610ff8db1eb8478ac5fd345
https://linux.die.net/man/3/bn_internal
https://www.microfocus.com/documentation/enterprise-developer/ed60/ES-WIN/GUID-E3960B1E-C42E-4748-A5EB-6E12507C9CD7.html
https://www.microcontrollertips.com/risc-vs-cisc-architectures-one-better/
Fundamentals of Networking for Effective Backends udemy course (link redirects to udemy with coupon)
https://network.husseinnasser.com
Postgres 15 introduces a new syntax to treat nulls as not distinct. This will allow developers to create a unique constraint on nullable fields and have inserts with nulls fail instead of having it allowed based on today's standard.
Fundamentals of Networking for Effective Backends udemy course (link redirects to udemy with coupon)
https://network.husseinnasser.com
Fundamentals of Database Engineering udemy course (link redirects to udemy with coupon)
https://database.husseinnasser.com
https://blog.rustprooflabs.com/2022/07/postgres-15-unique-improvement-with-null
https://www.postgresql.org/message-id/flat/84e5ee1b-387e-9a54-c326-9082674bde78%40enterprisedb.com
memchr() is a function that searches in a memory block for a character, this has been recently improved in the latest linux patch and we can learn something from this improvement I think, lets discuss.
https://www.phoronix.com/scan.php?page=news_item&px=Linux-Kernel-Faster-memchr
Fundamentals of Networking for Effective Backends udemy course (link redirects to udemy with coupon)
https://network.husseinnasser.com
Amazon Aurora PostgreSQL-Compatible Edition now supports PostgreSQL major version 14 (14.3). Let us discuss this news.
https://aws.amazon.com/about-aws/whats-new/2022/06/amazon-aurora-supports-postgresql-14/
0:00 Intro
1:00 Database on the Cloud options
3:45 Amazon Aurora supports Postgres 14
6:00 Postgres 14 vs Postgres 13
Rogers Communications, the Canadian ISP is down. One of its ASNs (they have many) is AS812 with over 5 million IP addresses. Tried a few and none of them can be pinged from the US (not sure if ICMP is disabled or not)
This is a huge deal for all Canadians and businesses affected. So sorry for every one who is affected by this, and kudus to all the engineers at Rogers working to fix this for the past what? 11 hours now?
If an ASN goes dark like a Facebook or a Cloudflare that isn't a big deal you can go without using Facebook for a day. But if this is your ISP's ASN that connects you to the rest of the World goes down, actual users won't be able to connect.
We still don't know the cause but my guess it might be a bad BGP entry? that's what happened to Cloudflare or FB, could be something different. again so sorry for all my Canadian followers affected. Will make a video when I get a chance
Resources
https://www.bigdatacloud.com/asn-lookup/AS812
https://www.bigdatacloud.com/asn-lookup/AS7018
From the publisher's feed

31,929 Listeners

274 Listeners

373 Listeners

12,159 Listeners

987 Listeners

8,070 Listeners

210 Listeners

1,626 Listeners

10,106 Listeners

550 Listeners

513 Listeners

5,565 Listeners

671 Listeners

1,439 Listeners

74 Listeners