Thank You for what you do

Level1Techs INBOX.deb
1

So at my day job we run a VDI environment and historically have always struggled with graphics-enabled stuff. From our clients, confidence has always been low in the viability of the deployment model. Most of that was bad prior leadership coupled with a scalability problem. We haven’t reinvested in this tech since 2015 when we bough 9 R730s with GRID K2s.

Fast-forward to me being a geek. After watching the most recent L1-Liqid collab, I arranged a meeting with Liqid with one of our VARs. At first they are like “who is that”–and then they are like “oh crap we know who that is we just started doing some big projects with them”.

As an aside, one of their SEs asked me later in the day what the meeting was about because he couldnt attend. I guess he had read the slide deck and didn’t understand fully what the offering was. I told him it was like having disk shelfs for PCI-Express and you could individually control what servers get them on a granular basis. He was like oh that’s cool I guess. Apparently not everyone appreciates the elegance of the Liqid solution…

Then I had staff members ask me “Wait you mean no one has ever done this before??” I just directed them to the interviews from Level1…lol

So anyway when I got on the call with them this morning the Liqid folks asked how I had heard of the company and why I had asked my VAR to look at them over our traditional (Cisco UCS) offerings. I started laughing because I realized I am wearing my L1 Techs hoodie (not planned) and I was like, so have you ever heard of this guy @wendell on YouTube, he has a channel called Level1Techs?

They were BEEEMING. They are like OHHH WE LOVE WENDELL and then they are like do you know that other guy, TechTips or something? I lost my shit laughing, but apparently they have a collab with Linus in the works.

In the end, the conversation was very productive and my staff and I all really like the product…and I hope to move forward with using them for a couple of projects. Apparently, they thought so too:
image

image

Anyway just wanted to say THANK YOU to @wendell for all he does. I don’t think there is any other single source of information or inspiration that has impacted me personally as LevelOneTechs (including OG TekSyndicate!). I mean that in both a personal and professional sense.

I am a proud Patreon subscriber, I love your merch. I just wish there was more I could do to thank you! I just hope this puts a smile on your face

15 Likes
2

liqid has managed to carve out a particularly interesting niche wrt. PCIe switching and “orchestration” … disk shelf for PCIe is a pretty good description actually.

Even if you’re doing ML workloads on k8s with GPUs as extended resources for containers, you can still do stuff like have GPUs for your desktop-y workloads (interactive CAD / physics simulations) during the day, and can attach them to k8s cluster over night, and also you can have a bunch of 100kW+ racks of GPUs in part of the building where you have thiccc power cables and high capacity cooling, and your mishmash of CPU dense stuff that’s not as power hungry in the “less hot” isle.

It maybe doesn’t make sense for hyperscalers, who are probably looking at this going. “PCIe is just remote ram reads/writes anyway with a funny wake-up/interrupt thing how’s this different from RDMA we already have and why don’t we just ask our CPU / PCIe switch vendors who we already have a good working relationship with, to support it mapping devices through our 100Gbps+ nics”… (I’m imagining this would be how this would go, no idea).