A fun story! Packet support contacted me one day last fall to ask about my business. I said I had none, I used Packet years before then for personal projects and only for a few days. Then they asked what my plans were for the 13 VMs I had running. I knew of no such VMs.
I guess my account got pwned.
They asked how I'd pay for these - $900 worth I think - since I didn't have a valid card in their system (the previous one expired since I last used Packet). I told them to stop the VMs and allow me to reset my credentials. I also asked them to investigate their logs and geolocate all the users that logged into my account. I was also curious they actually let someone log into the account from a new IP (without notifying me) and also allowed them to run instances without a valid card.
They blocked my account, which also blocked this support thread. So I started a new thread from an alternative e-mail address to resolve the issues.
It really depends on what you consider a major outage. If we set the bar at at events that would break customers using a single hosted location, 4 over two years isn't horrible. Things in this category are fiber cuts, automatic transfer switch failures and other power issues, PoP router mishaps, HVAC failures, etc. Steps can be taken by the provider to reduce the occurrence of these, but some will happen, and customers that care should build around it.
If we set the bar at events that break customers who have taken reasonable steps to be geographically distributed, 4 over two years is horrible. Things here are cascading router failure, BGP misconfiguration or external hijack, major peering dispute, bad push to critical systems. If these aren't rare, it's likely to be avoidable failures caused by the provider. Sometimes the earlier types of problems hit here too --- fiber cuts between regional data centers can be major problems if more things run on that fiber than expected; it's always fun to learn redundant fiber is in the same bundle, and usually that's learned when the bundle is cut.
same exact situation...we moved over all bright eyed with hip skip and a hop in our step...within 24 hours we experienced a full system-wide outage followed by a full system-wide block storage outage a few days later....their system has the reliability of a hobby project
I think the responder was questioning the phrase about packet asking about the customer VMs. You can certainly run VMs on top of bare metal, but packet doesn't know about that.
Likely bad news and writing on the wall for the open-but-competitive culture at Packet. Equinix has been historically turning every single acquisition into a closed-off enterprise-only product/service, especially in the IXP space.
I didn't know how fast to turn around when I tried to register an account and they insisted I needed to provide a copy of my passport or drivers license. I don't know why I need to provide such information for a hosting provider as I've never had to do so before at any of their competitors.
They obviously wanted to verify who you were which is why they requested a government issued photo ID.
Good on them for being proactive about it instead of waiting until you had created a server. Could you imagine a production server being frozen while they wait for you to verify your identity?
Fraud is incredibly prevalent in the hosting space. Spammers, CC fraud, hackers, etc.. all stand to benefit from servers not obtained in their own name.
> Fraud is incredibly prevalent in the hosting space. Spammers, CC fraud, hackers, etc.. all stand to benefit from servers not obtained in their own name.
All of those can get fake id scans for a couple of dollars each.
> I didn't know how fast to turn around when I tried to register an account and they insisted I needed to provide a copy of my passport or drivers license.
Hosting / pay for remote virtual stuff is notorious for abuse. I'm surprised AWS doesn't require an ID to get started these days. I understand why smaller player do, though.
hi I'm a mere frontend developer who doesnt do much at the network layer - do you mind ELIFEDing to me what exactly Packet got right? What's the big deal about automated bare metal? is it somehow a lot faster? cheaper? than virtualization? Is Packet 100% for edge computing or is Equinix kinda repurposing it in this acquisition for its edge computing strategy?
I feel like something that is so obvious to you is not obvious to me and it'd probably help other readers here to follow along. Thanks in advance
Here's the way I see it: there are a lot of customers who don't need all of the AWS services. They also would prefer to be running on bare metal for performance reasons, and to have slightly more control over their application. I know AWS with the nitro instances are closing that gap, but packet was around before they had those.
Also, their provisioning software is excellent. From the time you choose to deploy, it takes only about 30 to 45 seconds for your instance to start. I know AWS is faster, but AWS isn't installing a new operating system each time. as the other posters pointed out, they have a lot of features that AWS will also not let you do, like bgp with ecmp.
lastly, they tend to have newer types of instances that other providers don't that you can play around on. The cavium arm processor comes to mind, which I haven't seen on any of the major cloud providers.
ok awesome. minor follow up - why do customers care about the newer instances like cavium arm processors? again is it a pure performance thing? or they are extremely optimized for some kind of workload?
maybe i'm so used to OSes abstracting things that I completely lack knowledge about what baremetal choices can do for me. this is a little scary.
I can't speak for everyone, but ARM processors may be cheaper for some types of workloads. In terms of cores/$, they're cheaper, but some applications would perform far worse on those than on x86. Others just want to use them as a playground to test ideas without buying an ARM server.
Well, bare metal is usually way cheaper for way more performance at the cost of often being less managed.
What packet got right is that you get the same functionality that you could usually only get in colocation environments using high-end networking equipment.
You may want to look up on the Internet for “BGP” and “Autonomous System” to learn more.
ha funny enough i have done a networks 101 course so am familiar with BGP and ASes. just feel like thats a big leap to Packet which is much more cutting edge.
Vultr's BGP is not very good though. I regularly lose the full Internet BGP table from them and have to ask support to fix it. Supposedly it's some sort of limit of their network infrastructure. Another day they just randomly stopped announcing my prefix to the Internet, and since they don't even have a looking glass it was impossible to debug without, again, asking their support.
This is probably a good play for Equinix as it provides a lower barrier to entry to a segment of the market that they typically haven't engaged (Packet being largely profitable probably helps).
In addition, I believe Packet already had presence in a number of Equinix datacenters so "landlord buys tenant" comes to mind...
Yeah, it feels like it has a lot of potential on paper, packet does a lot of stuff well that Equinix probably was thinking "why aren't we doing that? why isn't that us?"
It feels like one of those cases where you really want that vision to survive at scale.
Making it easier for Equinix to leverage its resources to get people cloud-y style capabilities that isn't attached at the hip to the other clouds sounds like it would be a pretty good thing. And for Equinix themselves, being able to expand their direct offerings without stepping directly on the toes of the other cloud giants that are also its customers is likely always a good thing in itself.
I believe their pricing model was cheaper than any of the major cloud providers for an equivalent compute and networking resources. Given that their margins are also lower due to the volumes, I wouldn't think that it's obvious that they were profitable just by their pricing model.
I was hoping for Packet to be the next big bare metal provider so this is kinda sad at first glance. I’ve heard great things about them but haven’t used them; I was considering them for a side project of spinning up my own k8s cluster just because.
GCP and AWS probably fit the needs of most established businesses. I know quite a few startups that use bare metals though since they cost a lot less. Plus, you have complete visibility and control over your entire stack.
I hope others spring up to address the market of users needing extremely highly customized systems at the lowest costs.
Considering their service offering I understand it, but they don't have pricing I'd consider hobbyist-friendly. So building a skunkswork project out of your pocket is prohibitively expensive. Having said that it's one of the few places that gives you enough control over your device and network to let you stand up origami infrastructure where you build a system that builds some systems that have a kubernetes cluster that hosts your app.
Packet’s pricing doesn’t seem worth it to me compared to GCP - looking at their list prices they seem to be priced just 20% less for compute (actually same if you get e2s) so unless I’m missing something seems like very niche offering...
GCP will literally shut off your servers, with email support only, and f*ck you. Happened to me right after I'd done a multi-million job for Nike the weekend before. No recourse, nothing. Just "we think this is fraud so we're shutting you down, here's a noreply email to contact us".
I wouldn't consider using GCP for anything I cared about. AWFUL customer service.
They also offer discount pricing for long term contracts and etc. But depends on your use case among other things on whether a long term commitment makes sense.
Depends of course peoples needs for cloud. Some people need the power of bare metal and that's why most cloud providers are moving to making that an option.
Probably mostly for marketing purposes (just to say they have it). Kvm/nitro penalty is within single percentage points these days so 99% of folks out there won’t even know the difference.
"When the transaction closes later this quarter, Packet will continue operating as before: same team, same platform, same vision. Our commitment to our customers ..." -- The question is how long before Equinix decides to change the status quo.
I'm seeing it as yet another (after recent news on Visa's Plaid acquisition) blow to diversification and competition in the cloud and technology ecosystems. Generally, industry consolidation is IMO a scary thing ...
This is sad news. I used Packet for a benchmarking project (https://github.com/senderista/hashtable-benchmarks) and was super impressed by the ease of use. Put EC2 absolutely to shame (and was much cheaper than EC2’s bare metal offering).
I'm not sure that very many people understood what sort of spot pricing Packet had. I have rented so many hours of EPYC server time for $.25/hr in SJC1. Or a 96 core ARM machine for $.30/hr in Amsterdam.
I had good luck with them and they were always responsive in their Slack. I was sad to see the news, anyway.
What makes Packet any difference to, Linode, DO, Upcloud, Vultr, Hivelocity, and many others? Where the first few are bringing Bare Metal to its offering along with Database aaS.
Equinix has long been in the cloud game. Most public clouds have at least some presence in their data centers or using their global network backbone. I do believe this is their first venture into direct to consumer cloud sales assuming they retain Packet's business model.
I guess my account got pwned.
They asked how I'd pay for these - $900 worth I think - since I didn't have a valid card in their system (the previous one expired since I last used Packet). I told them to stop the VMs and allow me to reset my credentials. I also asked them to investigate their logs and geolocate all the users that logged into my account. I was also curious they actually let someone log into the account from a new IP (without notifying me) and also allowed them to run instances without a valid card.
They blocked my account, which also blocked this support thread. So I started a new thread from an alternative e-mail address to resolve the issues.
I never heard from them again.