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
The video explains a lot of it. Not sure why you think they don't use Anycast. Their entire infrastructure is designed to use Anycast.
But it's still proxied.
Francisco
How else do you think you would do this on your own servers with multiple providers?
Yeah still don’t think they do that. Nothing in the video seemed to say that they would. It seems no different then putting haproxy on a buyvm Anycast IP. I don’t think there is any reason to believe they do ‘internal’ anycasting, that’s just not a thing. But maybe they have some undocumented feature
If you don't understand all the differences then I guess you don't need it. So enjoy your haproxy + BuyVm servers.
I know that silly! I don’t think you understand what @gbshouse’s service actually does
I guess not. Even though I am using it right now. But that is also different from what totaluptime is doing so not sure what you are talking about.
Yeah I’m using it as well so I guess we both don’t know it works!
There is actually alot more to it than that for good Anycast services. For example:
Do we really want the Anycast provider doing this? I feel like everybody’s usecase is different and I can think of some good instances where you don’t want that
What would be the benefit of having the same upstreams/transits at multiple pops? Wouldn’t I want to peer with the providers that provide the least amount of hops and latency between the pop and the clients, which would depend on the location
Yes you most certainly do unless you have a large engineering team to do it in house. There are some applications (e.g DNS) where it's unnecessary but none I'm aware of that it causes issues with. The majority of uses cases benefit.
iBGP is always preferred over eBGP for balanced routes you need your international transit to be balanced (or have REALLY good traffic engineering communities). There are also other issues relating to local preferences of the next hop. All of which can result in longer transit paths. eBGP typically has no idea where your route originates.
For webpages that aren’t doing heavy downloads I can’t see much of a problem of a reset if it goes to a different pop. I remember this LinkedIn engineer article where they talked about this same problem and with all their testing route changes just don’t happen that often and when they did, just about every time they always came back to the same pop. So they don’t even worry about that problem.
And when downloading a big file from a CDN most times they do a redirect from the anycast ip to the pop ip range so if your route changes your still talking to the same pop ip
What I see a lot of the big providers do is do anycast per region which makes a lot of sense so you get the good local routes and not have to worry to much about international. They do dns geolocation to pick which anycast ip to give the client
@LosPollosHermanos feel free to contact me with questions/feedback
Is there an ETA for a production ready stable version?
@jimaek mid September
Small update for beta testers: some of you noticed routing issues today, had to reconfigure internal routing as cross region communities/prefs were ignored, it should be stable now. Added PoP in JP for better coverage in Asia, will add it to control panel tomorrow. Waiting for UK BGP session to be configured. Thanks for feedback.
Can we get a screenshot of the beta panel?
First impressions: deadly simple interface with little hassle getting setup (literally ~4 steps total per machine), very interesting use of ZeroTier as an overlay, no need to dick around with tunnels or any of that for back-haul.
Still need to do a bit more performance testing and run some HA/failover scenarios but very impressed thus far.
Below is a snapshot of ICMP response times as seen from Oracle Internet Intelligence. <50ms from major markets, 50-99ms everywhere else, with a few outliers.
How are you doing HA/Failover? I can't get that to work. I have to manually remove the anycast route for that server from the control panel before it goes to a different region. Just shutting down the server (which simulates a server/datacenter issue) doesn't do anything. It still tries to connect to that server.
@LosPollosHermanos - I don't remember mentioning anything about HA/Failover
it's not yet officially there. Still working on this bit, have few ideas but need to test them first
Quick update for all beta testers - just published refreshed version of control panel, feel free to log in and take a look. Routing in London will be fixed tomorrow (Monday). @LosPollosHermanos added service which monitors the state of each ZeroTier endpoint and adjust the routing, feel free to test it.
Very interested in using this, unfortunately BuyVM doesn't have an Oceanic location so I haven't been able to use them - PM Sent!
So couple of UI screenshots
I prefer to keep UI as clean as possible, feel free to comment.
Looks good! Still waiting to hear about the pricing though.
Also, nice domain you got there.
@vovler we like the domain too. Still not sure about the pricing, will have more info by the end of next week. Currently we have Sydney, Singapore, Tokyo, Los Angeles, New York, Amsterdam and London. Sao Paulo is waiting for deployment, the same South Africa. Before the end of beta will also add couple PoPs in US and EU.
Any news on it yet?
So like x4b without the ddos protection part? This works for any applications right? Interested in trying it out to see what happens.
@vovler sorry for delay, we've been busy upgrading our DNS platform, pricing will start from 25€/month with 1 IPv4 and 1 IPv6 included, not sure about traffic amount
Is this out of beta yet? Nothing on your website.
nope, we've decided to do beta till the end of the year