New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
^this
That's not true at all. I have redundant VPSes across different vendors across the planet. Short of a software bug (which can also be mitigated), for my system to go down ALL of the providers would have to go down along with ALL of the internet and ALL of the electrical systems, etc. As long as one is still standing I'm still at 100% uptime.
You don't seem to understand distributed systems.
Read the above.
I understand just fine.
Exactly. Probably far more possible scenarios that could effectively bring you down than you'd like to think too. Registrar? DNS? The system controlling the failover distribution? You can never eliminate single points of failure completely. You can never guarantee 100% uptime. There is always something that can go down and reduce you to 99.99% like the rest of us mere mortals. You can reduce probability, nothing more. As I said, 100% uptime is about best effort and a dice roll. Nothing more. So if 100% is the only acceptable number to not be "crap" then that's fine, but it's not just skill and planning, it's luck. So you can never have anything but "crap" unless the creek doesn't rise.
@nutjob worse stuff has happened:
http://blog.cloudflare.com/todays-outage-post-mortem-82515
One small detail can take a good bit down.
You never plan for it to happen. Just try to be prepared the best you can.
Find and tie Murphy.
Have you also rewritten your business plan to allow for this?
Yes. And people will also pay $20 for 1GB if the host is capable and truly delivers a reliable product.
If I want a VPS for a personal VPN for whatever use, then sure a dirt-cheap service for a couple bucks from the latest startup is fine. I can get two or three -- one or the other should work when I need it.
But if I want to host a website and aim to have to up 99.5%, then I'll shop around and I'll be willing to pay much more. I'll want to do business with someone I feel I can trust (that include business integrity and technical competence). I'll expect to pay in proportion to what I need. The GBs and MBs and MHzs aren't the deciding factor, it's the reputation.
A host with a solid reputation for quality of service doesn't need to compete for the bottom-of-the-barrel market. They will come.
While I agree that there are likely single points of failure that he did not take into account, I don't think that it is inherently impossible to eliminate single points of failure - just improbable. Right now, the reason DNS is a problem is because you can't reasonably do without it and you can't just replace it either (for compatibility reasons). It's a single point of failure that's mandated by one or more third parties.
Theoretically, if you switch to an entirely distributed DNS and registrar system (they do exist, just not widely used), that would take away that single point of failure. Theoretically, assuming all your nodes can be considered 'trusted', you don't need one single system controlling your failover behaviour, and you could have all your systems controlling failover behaviour, thereby providing redundant failover handling.
All of this, while possible in theory, is improbably in reality because you're stuck with certain bits of infrastructure that you cannot work around without breaking it for a lot of people, or introducing some other undesired circumstance. That does not, however, mean that it's impossible to eliminate single points of failure. Just improbable and a lot of work.
@sleddog no I just adjusted the business plan a little bit.
Very aggressive prices will mean you have big chance of having bad neighbors
Nope, nope, nope and nope. As I said, you don't understand distributed systems.
IP numbers are distributed by the system and multiple seeding domains/IPs are with multiple providers, so there is no single point of DNS/IP/routing risk.
There is no failover mechanism. Each client talks to multiple servers, servers talk to multiple other servers and everything organizes itself.
As I said, the biggest risk is software error, but even that can be mitigated.
Keep your day job, Jarland.
So you're saying that your domain will resolve no matter what point goes down and you have multiple A records none of which are recorded in a single destination and no setup that will require a few minutes downtime to the average visitor if any point fails? You use no domain names for services? Your "clients" have no single location to query at any point?
You're not immortal, you're just being so obscure about what in the world you do that I can't pin point your potential point of failure that will drop you below 100% uptime. If what you do has no human client facing portion then it's irrelevant to the rest of us and it's nothing more than a node circle jerk. Unless you're handing out a list of IPs and telling your clients to visit each one until one works, you've got a single point of failure. Sure it can be self healing, but downtime is downtime and as you said only 100% is acceptable.
Try not being a dick?
I agree that human error and software error (which is human error) can be a problem, but that can be mitigated. I can go into that if anyone wants to hear it but I'm sure I bore everyone to death (if I haven't already).
@nutjob I really think you are just arguing for the sake of proving everyone wrong and you are right...seriously, I failed to see the point when you are arguing about "this is LEB and we should talk only about cheap boxes" at one point and then moved on to talk about "distributed systems (which are by no means LEB)" the next.
Ugh. Reread what I wrote. Then do it again. Where is the "domain". Multiple domains are queried in each client for multiple candidate server IPs. No single point of failure.
I'm just following your lead, Jarland. You have an attitude about a subject you obviously know nothing about... because you administer a few servers? Which is fine, you can do say what you like, but don't whine when it comes right back attcha.
But in the name of civility, I'll tell you what: I'm going to ignore you here on in, because this is pretty unproductive.
I'm explaining how I'm using the LEBs and how making them cheaper even if less reliable works for me and might work for others. I'm illustrating my point. What exactly is wrong with that?
Is it just me or do all vendors have attitudes?
All I can say is...LOL
Are you @bamn?
Read:
Now...
Coming from the guy who does...what exactly? My qualifications are not based on Catalyst. Admittedly it has helped me understand infrastructure better in real world environments. I'm just asking for clarification on how you've transcended reality and obtained absolute perfection while the rest of us mere mortals are only able to obtain reduced probability. If you wanna cry like a little child and hurl insults then good day. I'm talking technical, this isn't high school. Grow up and converse like an adult.
You're really not though. You're just saying you've obtained 100% internet perfection and a flawless infallible infrastructure with no details as to how that is remotely possible.
The point is you are arguing over something that quite frankly speaking, no point for arguing. It is like arguing whether chicken is better than fish/ is KFC the best food in the world.
Again, if you want to argue for the sake of arguing, then you can do that at home instead of wasting everyone's time on a public forum.
+1, somehow I smell @bamn is coming back
Boys boys, now hug each other.
Probably lol
Honestly though I'm just trying to get a solid answer. If this guy knows how I can keep my website running 100% and be completely shielded from any and all variables that could possibly change values due to natural events or human error beyond my control, using LEBs, he's holding the answer to life and I want it.
+1, I guess @nutjob can enjoy his own sweet time trolling around then.
Let me know as well. Because then I'll be rich selling those HA LEBs