Slashdot Mirror
Linode Under DDoS Since Christmas (linode.com)
hol writes: Linode has been getting hit with DDoS attacks since Christmas Day, and it looks like their pain is set to continue. The attackers are rotating DDoS traffic through various regions of Linode's service. They say, "All of these attacks have occurred multiple times. Over the course of the last week, we have seen over 30 attacks of significant duration and impact. As we have found ways to mitigate these attacks, the vectors used inevitably change. As of this afternoon, we have mostly hardened ourselves against the above attack vectors, but we expect more to come. ... Once these attacks stop, we plan to share a complete technical explanation about what has been happening." See their status page for updates.
WTF is "Linode"?
I've used a variety of hosting providers, but I always keep coming back to Linode. Their product is competitively priced, they provide exceptional service and support, and they are very simple to use. And, unlike AWS, you don't need a calculator and 2 hours spent parsing fine print in the documentation to figure out how much a given level of service will end up costing you. I highly recommend Linode for your cloud computing needs. I hope they are able to resolve their DDoS problems quickly.
Null routing an IP address under DDoS attack in an emergency is standard industry practice across all major ISPs and hosting providers; companies that use more advanced techniques either have a few tricks up their sleeves which only work in the most common situations, or they bought some $5 million anti-DoS appliances to help mitigate it (usually).
The simple fact is DoS mitigation is not part of a basic hosting service, once an attack exceeds a few million packets per second, or a couple Gigabits: you are simply not paying network providers enough money for it to be feasible for any ISP to come close to justifying effective DoS mitigation for those rare sizes of attack, for every customer, because the cost involves provisioning hundreds of million$$$ in extra upstream capacity, internal network capacity, and operations staff.
Then even with all that extra capital spend: (1) It's still not possible to make every attack seamless, Null-routing might still be required in cases, there will still be outages, people like the above will still be unhappy, And.... (2) Most ISPs don't have that much throwaway cash, and most hosting customers aren't going to be willing to pay their share of what it costs to provision 10000x as much capacity as needed.
(3) Its less expensive to just shed overly-demanding customers who pay little by allowing them to make themselves unhappy and go to a competitor. If someone's paying $100 a month and their site is constantly getting DDoS'd, then it makes perfect sense to terminate them as a customer to, and let the other 10000 $100/Month customers have a better experience, instead of leaving due to the DoS being suffered as a result of 1 customer.
And if someone wants to arrange for their website to be handled differently, then this is part of a negotiation that should be made with the ISP or provider before turning up hosting service and added to the contract, with response SLA and recourse/refund policy.
Or you're better off enlisting a 3rd party DoS-scrubbing service such as CloudFlare to conceal your infrastructure from attackers.
There are also DoS-cleanup services that work at a network range level where your DDoS provider announces your /24 into BGP, cleans DoS, and forwards you traffic.
Many ISPs do have the flexibility for alternate handling of DDoS, up to a certain point, they can avoid Null-routing an IP, or avoid the Null-routing of one IP from making your service unavailable.... generally, the cost will be much higher --- E.g. $10,000 per month instead of $100 per month.
Forget about attempting to negotiate expert-level DoS management that will require the provisioning of engineer and infrastructure resources in advance that are quite costly to the providers to keep on hand, Unless you are willing to pay sufficiently to be a large client of the provider with a multi-year committed contract and cover the costs of those extra resources plus sizable profit.
Also: to host a website resiliently, however, the provider will most likely require that the website be served from multiple server farms in multiple IP ranges with an anycasted internet presence for both the services' IP addresses, and the supporting DNS services.
This is because in spite of additional resources, it might still be necessary at times to fall back to Null-routing.