But how does it attain consensus? No real decentralized platform can reach that at the present technology we have on blockchains and distributed ledgers. Mining1 is right it is centralized. It is not the same banana with many of the cryptocoins here.
Ripple's technology can. The test used geographically diverse validators. The validators could have been operated by any combination of organizations. A live, decentralized network would have reached almost exactly the same limits.
Unlike mining, the consensus process isn't significantly affected by the number of transactions per second. Validators propose groups of transactions by the hash of the tip of the Merkle tree that includes them and other validators do not have the perverse incentive of gaining a benefit by locking out other validators.
I don't see a point here, 1000 transaction per second is not world record or anything. Bitcoin is much slower, true, but bitcoin has store value potential Ripple hasn't.
Centralized online money payment processors have much more capabilities than mere 1000 tx/s. VISA has a peak capacity of around 56,000 transactions per second, beat that first.
That's not really an apples to apples comparison. Sure, if you don't want a decentralizable, public ledger system then you can build private, inherently centralized systems that are faster. I don't think we'll see, for example, high-frequency trading on a decentralized public ledger in the foreseeable future.
Also, if you're going to compare Ripple to systems like VISA, you should compare Ripple with the additional layers that scale it such as payment channels. And when you look at scaling, it's essentially a multiplication of the base ledger's capability -- and it matters what number you're multiplying.
But I also think there's another point that's important. Nobody needs 1,000 transactions on a public ledger today. When you have way more performance than you need, you can trade it for other things that you do need.