Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Joyent Public Cloud EOL

docs.joyent.com

Post reply on HN
Print view View original

91–100 of 289 posts

Re: Joyent Public Cloud EOL

#91

Post by Thaxll »

Annatar wrote: post #24
Thaxll wrote: post #21

Earlier quoted context omitted.

Yeah how can you compete against AWS / x86 / Linux on that space not sure.

With technology which is an order of magnitude more concerned with reliability and easier to use than any competition out there, that's how.

By using Solaris / re-implementig KVM on different arch / platform, ofc it's the wrong approach, on one side you what 100s eng at Joyent working on that, where the rest of the world uses native KVM, x86, Linux ect ... it's just not possible on a tech level to compete.

Re: Joyent Public Cloud EOL

#92

Post by ForgotMyUName »

morning_gelato wrote: post #69
sean0- wrote: post #61

Earlier quoted context omitted.

The FreeBSD/VPC hypervisor efforts ran into a few “corporate speed bumps.” A few folks, myself included, are working to merge this code and make it generally available. The FreeBSD hypervisor effort certainly was the best systems work I’ve seen in my life, especially in such a short amount of time and given the resources at our disposal. My favorite anecdote was that the internal first-pass review of the benchmark da…

Wow pushing 100gbps on half a core without hardware acceleration is an incredible feat. Can you provide more information on how this was done?

Well you see, Sean is a known liar and conman. So it wasn't actually done. It is one of the reasons him and his team were fired by Samsung when their house of lies came tumbling down.

Or uh, so I heard.

Re: Joyent Public Cloud EOL

#93

Post by rjzzleep »

bogomipz wrote: post #83
cbsmith wrote: post #53

Earlier quoted context omitted.

Joyent was a public cloud provider, that offered a VPS/Container style offering built on Illuminos (open source Solaris). The expanded into offering Linux using some clever stuff to support Linux runtime on Solaris (kind of like WINE, but much simpler since Linux is much more similar to Solaris and also fully open source). All with DTrace so you had some great transparency on what was going on with the underlying har…

I thought Joyent was built around SmartOS? Although I never really understood the significance of the different Open Solaris forks. Or are these the same?

Disclaimer: I have a smartos setup because for now it still runs quite stable.

Joyent/SmartOS was the one that added KVM support and a bunch of other features to the kernel, but I think it was merged into IllumOS. It's quite limited and has a clunky interface. When I tried to contact anyone there to add UEFI boot support there wasn't even so much of a: "we don't care".

The problem with SmartOS, Illumos and the rest of the Solaris world is that while for Linux you can find heaps of books, forums, stackoverflow entries there are 3 books for OpenSolaris. SmartOS has seemingly great design, but it's absolutely nontransparent on how to use it. The documentation is subpar, lots of the wiki reads more like an experiment than an actual documentation.

If you hit a bug, be prepared to spend hours on github issues to see if someone found a solution to it(might as well skip it because it's unlikely).

If you use Triton and the API for it all is good, but if you want to use a SmartOS cluster directly, everything thing will be fine until you make a mistake or try to do something non standard and then good luck to you.

There's a great/fun talk by Bryan cantrill about how management killed Solaris. But if you ask me it's only one half of the story. The other half is elitism.

If you want people to engage in your open source project it needs to be decently reasonable to work with and you should be able to get some sort of feedback, even if that feedback is "lol, what you're doing is completely wrong why don't you RTFM here".

But the attitude I saw was more like: "Our stuff is greater than everything else so why should we bother with you".

Re: Joyent Public Cloud EOL

#94

Post by saurik »

One nice thing about having more cloud providers is having more options for where your computation was housed. I have had a machine at Joyent for a long time now (with the goal of building a service that I never got around to, so sadly for them only ever the one computer) because their US South West datacenter is located not just "in Vegas" but specifically at the SwitchNAP SuperNAP site, which co-located me with other resources I was using: I had a <1ms ping to my upstream telecom provider, and thereby could sit inside of real-time audio without adding noticeable latency (as well as have the option of getting extra-cheap bandwidth: I wasn't going to be costing Joyent anything for bandwidth but was still intending to use lots of CPU, so they had indicated a willingness to let me one-off this billing if I ever actually scaled up, though I was still quite happy to pay the full price for the epic latency). AWS is great (and I honestly used them for most of my less latency sensitive projects... hence the problem for Joyent, I appreciate), but their lack of a South West location means that the lowest ping I can get from them for this purpose is almost 20ms :(.

Re: Joyent Public Cloud EOL

#95

Post by snaky »

hinkley wrote: post #89
alynn wrote: post #64

Earlier quoted context omitted.

> My understanding is that Joyent is comprised primarily of Sun refugees This is by far the best explanation I have heard. Suddenly Joyent makes sense to me after all these years. They have been continuing to do what Sun did. In a very similar way to how Sun followed the fatal strategy of simultaneously trying to make money selling proprietary Unix hardware, while also making their software open source and porting it…

Part of the overengineering in J2EE comes from essentially trying to make one app server be multitenant instead of just spinning up one process per tenant. That this came from a company essentially selling mini computers. In fact, we are trying to do multi-tenant all over again. To my mind, this cycle of containerization is still trying to fulfill all the promises we were given in the early 90's about memory protecte…

Didn't AS/400 fulfill the promise? OTOH it may been considered 'mini-mainframe'.

Re: Joyent Public Cloud EOL

#96

Post by kbenson »

rjzzleep wrote: post #93
bogomipz wrote: post #83

Earlier quoted context omitted.

I thought Joyent was built around SmartOS? Although I never really understood the significance of the different Open Solaris forks. Or are these the same?

Disclaimer: I have a smartos setup because for now it still runs quite stable. Joyent/SmartOS was the one that added KVM support and a bunch of other features to the kernel, but I think it was merged into IllumOS. It's quite limited and has a clunky interface. When I tried to contact anyone there to add UEFI boot support there wasn't even so much of a: "we don't care". The problem with SmartOS, Illumos and the rest o…

Not a IllumOS/SmartOS user, but can you comment on how good the man pages are? What you state is sometimes a complaint I hear about OpenBSD, but from my experience using OpenBSD (which was mostly almost a decade ago now) is that's because they actually put a huge amount of effort into making sure the man pages were not just correct, but comprehensive (to the point of requiring code patches to include an accompanying man page patch if it added/changed a feature or changed functionality).

When first using OpenBSD this was jarring, because I just wasn't conditioned to expect that system level features would have useful man pages, as that was almost never the case in Linux.

As a simple example of this, here[1] is the OpenBSD man page for ifconfig. Note that it contains liberal references to other man pages of similar quality and completeness where a protocol or functionality might be better explained. Once I learned this, very little outside help was required to setup OpenBSD firewalls stateful failover of firewall connection state as well as VPN tunnels with automatic IP failover with CARP. All from a handful of man pages.

So, that said, does OpenSolaris have good man pages? If it does, it might be that RTFM is the correct initial response to most questions, as it would be the up to date and definitive information about the what and how of things. A simple comparison of ifconfig[1] makes it look somewhat good, but that doesn't mean it's kept up to date and accurate. Can anyone comment if this is the case?

1: https://man.openbsd.org/netstart.8

2: https://illumos.org/man/1M/ifconfig

Re: Joyent Public Cloud EOL

#97

Post by bdha »

sean0- wrote: post #61
wolfspider wrote: post #47

Earlier quoted context omitted.

Yes I'm interested to know if this affects what they were doing with VPC control plane for FreeBSD and the subsequent integration with Vagrant and VMWare.

The FreeBSD/VPC hypervisor efforts ran into a few “corporate speed bumps.” A few folks, myself included, are working to merge this code and make it generally available. The FreeBSD hypervisor effort certainly was the best systems work I’ve seen in my life, especially in such a short amount of time and given the resources at our disposal. My favorite anecdote was that the internal first-pass review of the benchmark da…

"Corporate speed bump" is not how I would characterize any of what you did. Quite the spin!

I mean, if you want to publicly comment on your time at Joyent, please! I'm still pretty angry about it, and I left the better part of a year ago.

Re: Joyent Public Cloud EOL

#98

Post by throw0101a »

hinkley wrote: post #87
throw0101a wrote: post #72

Earlier quoted context omitted.

> My understanding always was that they're some special kind of hosting provider. But I never really got it. The main selling point is/was being closer to the hardware and thus getting better utilization. Virtualizing at the machine level was a lot slower (especially before Intel VT-x and AMD-V came along) and so one wouldn't lose (say) 30% right off the top: * https://en.wikipedia.org/wiki/OS-level_virtualisation

You will also hear a lot of bragging about how much more robust the partitioning is on Solaris versus Linux. Which is why they did the work to get Linux containers running on top of Solaris instead of just doing Linux.

I haven't been paying attention recently, but last time I checked, there were no known escapes out of Solaris zones, while there are quite a few CVEs about escaping Linux containers and even QEMU/KVM/Xen.

The one escape out of FreeBSD jails that I'm aware of (CVE-2005-2218) involved not a bug in the jails code, but by burrowing out via devfs.

So it seems that OS-level virtualization/security has a better track record.

Re: Joyent Public Cloud EOL

#99

Post by elijahwright »

Thaxll wrote: post #91
Annatar wrote: post #24

Earlier quoted context omitted.

With technology which is an order of magnitude more concerned with reliability and easier to use than any competition out there, that's how.

By using Solaris / re-implementig KVM on different arch / platform, ofc it's the wrong approach, on one side you what 100s eng at Joyent working on that, where the rest of the world uses native KVM, x86, Linux ect ... it's just not possible on a tech level to compete.

Joyent has had nowhere near hundreds of engineers, I'm sorry to say. Small tribe with amazing results.

Re: Joyent Public Cloud EOL

#100

Post by tastroder »

httpsterio wrote: post #73
randylahey wrote: post #42

Earlier quoted context omitted.

Ryan Dahl created Node.js prior to being hired at Joyent, and then sold them the rights to it.

The iojs debacle was because joyent had too much governance of the project and it was subsequently handed over to the board that currently governs nodejs. While the first Iines of what became nodejs weren't written by joyent, they can be quite reliably be credited to being the company that made it happen originally.

They recruited Dahl (and certainly contributed much through people like isaacs (npm, iirc he was at Joyent already) and more before he officially moved to Joyent) but if my memory of that time isn't all too blurry it certainly gained enough steam at that point to credit him as being the one kicking it off. It surely was the node we know today at that point (granted, with a much less enjoyable API surface). Not that this belittles their role, they certainly grew and provided a viable ecosystem for the large scale adoption that followed.
Post reply on HN