Wow. I can't imagine the thought process of going from accumulating a decade of Microsoft shares and experience during a pivotal era and deciding to plow it into a consumer scam op.
for some reason, the tech industry attracts these types like flies to...well, you know the saying. They aren't the majority of folks drawn to tech but outside of maybe finance, it seems to be their preferred habitat.
It's easier to see in hindsight but servers have eaten a lot of macro economic headwinds.
1999 and 2002 correspond roughly to telecom and dotcom bubbles popping. So open source operating systems were also pulled up into these positions through a lot of work and by the early 2000s it was kind of obvious this was the way forward.
2008 financial crisis really knocked the high end (RISC/UNIX/Itanium) and started the long migration to renting vs owning. In hindsight it's remarkable how much money used to just be there in the economy of all this. But I guess it still is, considering how expensive public cloud is.
> But I guess it still is, considering how expensive public cloud is.
It's not like there's less servers being sold, they're just bespoke designs from companies like the big Taiwanese ODMs that go straight into hyperscaler facilities, and we don't get to see public marketing websites and product info about them. The open compute platform stuff that's been publicly shown is like the tip of an iceberg.
> The open compute platform stuff that's been publicly shown is like the tip of an iceberg.
I find it a real shame it's not easier to get OCP hardware for our homelabs. Would be so much easier to operate, upgrade, and so on. And where would you even get decommissioned OCP gear?
OCP was designed by vendors, including ones that didn't want to get sunsetted. It was never going to be practical for homeless until EOL hardware started to filter down.
Some of the ethernet switching top of rack 1U stuff used to show up on ebay. I think large other portions of it go into an industrial shredder/recycling process.
I'm referencing the premium you pay to get some paltry VM in a public cloud with bizarre storage and networking that primarily make sense for the market maker versus the actual cost of the hardware. There are ways to make cloud economics work as a user, but only where staffing costs would dwarf spend (i.e. small businesses) and I've never seen them in practice.
Dell puts a fair amount of engineering into their servers (as do other main brands like HPE and Lenovo) so it's unfortunate they've lost their roots from the old mail order sales company.
If someone else is paying I'd prefer one of these big OEMs for established firmware lifecycle, spare parts and long gray market availability, shock (shipping) and thermodynamics engineering. If I'm paying, Asus or MiTAC (Tyan) are good enough.
I was fortunate enough to work with a fair bit of sun gear (3/260, 4/670, enterprise 250/450 and even some big iron 6500s) back in the day.
These days, I've got several Dell 5820 style workstations and I'm pretty sure someone who interned at sun in the 1990s is now running dell's workstation division. The medium sized dell deskside towers have nearly identical vibes as those old workstations, and are a pleasure to work with.
AFAIK those sold relatively well, considering the TCO of both sides (NICs were very expensive and there were like a dozen companies vying for position once upon a time). They were also interesting because they were cut-through. CX4 was noticeably lower latency than SFP+. So a very fast, but feature limited architecture which is probably why other players like Cisco and then Arista (aka Broadcom) took over.
You can, to some degree, do an online defrag with zfs rewrite -P now although it is not the primary intent. The displayed fragementation may not go down or may even increase, but all the files will go through the allocation policy. zfs fragmentation display is a little misleading, it measures free space fragmentation.
I don't think there is a ton of overlap. tokio is appropriate for general userspace apps, ranging anywhere from a CLI, GUI, API or web app. DPDK and SPDK are specialized fast paths for building network data paths and storage solutions that come with tradeoffs: DPDK uses poll mode drivers, outside of the operating system, which have various implications including busy waiting and taking over the interface. That is why DPDK is fast, no kernel/userspace context switching and copies, and the drivers are tuned for the polling model. But it's not a general purpose building block.
Fwiw with ef_vi you have full control over the event queue - you don't need to busy-spin it, you can choose whatever strategy you prefer.
> tokio is appropriate for general userspace apps
Yep, and for those I wouldn't recommend it. But tokio is also widely used in performance-critical infrastructure and web services. For those I'd say it can definitely be worth taking a second look at kernel bypass.
ef_vi is a solarflare proprietary feature which is now a support product, AMD moved on to Pensando. Once you move away from busy poll, you rapidly lost grounds to use DPDK. The PMD is a deliberate design to elide latency and lower interconnect taxes like PCIe traffic and cache/memory bandwidth by batching queue maintenance, that is the bargain made with a PMD. The field opens to OS native fast paths which have fewer downsides outside of that niche. Application developers are rarely concerned with this because it's far from where the bottleneck is for them.. a web service is rarely primarily a data mover, while a proxy is. Tokio has more in common with Golang than something like DPDK.
Definitely an accomplished architect, there are a lot of fine construction details in these photos. So much that it makes the IT/low voltage installation look like a hack job.
Corporate IT is kind of a parallel universe to me, so it is interesting to see how widely used these products are and how impactful these types of events are.
There are a few broad categories to building one of these platforms.. core virtualization ("relatively easy"), device model ("medium hard"), storage management ("hard"), management plane ("medium hard").
VMWare popularized x86 core virtualization and defined their own device model because they were the early market leader. But the device model is pretty bizarre and I haven't been particularly impressed by the guest driver quality over the years. VMFS and vSAN seem good for storage, the latter being better than anything open source. Management plane was mediocre, I remember it took a while to move off Flash and fat clients while cloud vendors were showing none of that was necessary from inception, and everything is kind of in a walled garden by design (embedded DB etc).
I had a lot of positive experiences with VMware (ESXi + vCenter). Every time I had a new project, I tried to keep it as simple as possible and it worked wonderfully. SAN, Distributed Switches, Automatic VM migrations to balance CPU usage etc.
VSAN was a weird one until I understood the logic behind it, but the licensing costs kind of brought on-par to buying a SAN and be happy.
reply