Simon Peccaud

Quick update: Since Sep'25, we have seen new kind of DDoS coming from US and South America (Bresil, Chili, Argentina, Maxico, Columbia). The size is around 15-16Tbps coming thought Miami,FL Dallas,TX and Los Angeles,CA. Last months, last weeks, we added lot of new anti-ddos (VAC) capacities and lot of bandwidth. Today we have >70Tbps anti-ddos capacities (VAC) and we continue adding 2-3Tbps (VAC) every week. Also, we have now >120Tbps bandwidth between OVHcloud (AS16276) and Internet (transit, peering, pni) and we continue adding 3-4Tbps every week. The plan is to have >100Tbps anti-ddos (VAC) and >200Tbps bandwidth. We continue working 24/7 to get there asap.

A brief retrospective of network-layer DDoS attacks in 2024 at OVHcloud blog.ovhcloud.com/a-brief-retros…

800G receiver chip (2x FR4, you can see two big arrayed waveguide gratings (AWG) for the demux) and 400G transmitter chip (four SOAs die bonded, one AWG, and a ring resonator? Fast modulation signals coming from the left, and bias signals from the right). From Super Photonics Xiamen, acquired my POET recently.

ghostty.org 👻

En 2024, les DDoS font 2.5Tbps de traffic entrant sur un réseau. Sous peu, ils vont dépasser 10Tbps. Il faut continuer d’investir dans les capacités pour être capable de recevoir ces DDoS puis continuer à investir dans les technologies pour aspirer et nettoyer tous ces packets IP inutiles. Chez OVHcloud, nous avons décidé de developper nous-mêmes toutes ces technologies. On les appelle VAC (de Vaccum aka aspirateur). « La team de tueurs » (les mots sont faibles) qui développe ces techno utilise les FPGA ou le DPDK sur les serveurs X86. Tout simplement. Ceci nous permet d’augmenter les capa de nettoyage pas cher et rapidement. Article de blog sur qq details de DDoS qu’on reçoit. blog.apnic.net/2024/11/07/the…

After months of intense work, the first article is finally out: explained-from-first-principles.com/internet/

I kept my mouth shut when SR Linux users inquired about our NetOps Development Kit (NDK), and I had a good reason, IMO. It required a developer to write an NDK app, and I wanted to simplify the NDK framework so that a network engineer could write those apps. Recently we finally devoted some time to NDK simplification, and it looks promising 😊 Yesterday morning I wrote this little NDK app that wraps sshx.io service and provides a collaborative SSH web terminal from the SR Linux CLI. But the beauty is that it is instrumented from the same SR Linux datastore, so you configure your SSHX client the same way you configure BGP. It is part of your NOS config tree, no less. You just install the (deb) package on your NOS and it merges into the NOS' YANG; now you can configure it, read its state, and operate it via any API interface. Simply beautiful. I'll do a proper video explaining how it works.

@fcuiller @MrRDGrey Il y a un micro contrôleur derrière qui sort sûrement du XOT yes.

@MrRDGrey @SimPeccaud Hello, it's 2024: X.25 Over TCP (XOT) Over IPv6 #nextGen

Rocking minitel tonight

GitHub Codespaces, along with tools like NetLab and ContainerLab make it easy to create and share labs. netautlabs.net has the goal of sharing network automation labs open source. Hopefully, many more labs are coming soon.

The Rise of Packet Rate Attacks: When Core Routers Turn Evil blog.ovhcloud.com/the-rise-of-pa…

Kudos the Team ! « The OVH's Optical Core Backbone has successfully transitioned to FlexGrid ROADM gear. This upgrade will significantly enhance our optical network's resilience, performance, and scalability.   This migration which took about 16 hours spared in two intense nights, could be considered as one of the biggest interventions applied to OVHcloud’s history, affected 264 optical channels (2 x 132) and 16.6Tbps of traffic without any customer impact.   The migration of Legacy to FlexGrid node, concluded 260 Active and passive optical elements (cards, modules chassis, frames, etc,) installation and almost 400 rounds of wiring on 4 sites.   With the Flex implementation, we now have the flexibility to enable 200G and 300G optical channels. Also now we will connect additional dark fibers in our optical backbone to manage a better resilience on the L1 and lower latency between our DCs and POPs in Europe »

@ntdvps @carlrmontanari @clabernetes Stickers ?

I think this gem of a logo needs to be printed out. This is for @carlrmontanari custom non-VXLAN datapath for @clabernetes

@bgptools Thanks BGP Ben for the alert! It turns out that prefix-length checking does contain some hidden man-traps

I admired your dedication to making networking accessible on all counts. May you rest in peace Nick.

In FY24, We have done : Our first 400G Peering port Our first 400G IX port And now, our first 400G Transit port Thanks to our team, our partners and suppliers. FY24 is not finished yet, more to come 🚀📈

🎉 OVHcloud's network has just expanded to Vienna, joining BIX IX ! Enjoying faster, more reliable services and a healthy dose of schnitzel-fueled connections 🇦🇹

@acontios_net Totalement 😼😂

@SimPeccaud Tu dis ça parce que je suis masqué ? Goujat !

Une bonne soirée 👍

@acontios_net T’es pas mal sur celle du milieu.

J'ai eu un veto pour la dernière photo... cc @SimPeccaud

I've won the 33.6 seat. Not even a 56k speed 😏 Bon appétit #RIPE88

Lajkonik dance & music group at #RIPE88 dinner @ Krakow