The simple answer is because of cost! The yellow line on the graph below illustrates why traditional cloud computing has been so successful: as you add servers, nodes and VMs etc, the cost per server remains constant. Since servers handle a certain amount of customers or users per serve, you can therefore conclude analogously that your costs per user stay constant with traditional cloud computing.
So why is the future of cloud computing decentralized? One of the greatest benefits of Sparkster’s Decentralised Cloud is that as we add nodes to the network not only does the TPS increase. Since prices of renting your spare capacity are dictated by the owner of the device, and not by us, there’s a natural competition to drive more volume by reducing prices. This competition will drive down costs to a point that is far below traditional cloud computing, so low in fact, that will disrupt the industry. How is this possible? Keep reading.
In last week's discussion we talked about a washing machine and the three pieces involved for it to function. The microcontroller controls the actual physical machine and a virtual function in Sparkster’s platform that enables you to create a routine such as a wash cycle.
We discussed how a delicate wash uses functions to operate, this involves the following sequence of events, or a routing:
-Turning the hot water on
-Wait 20 seconds
-Turning the water off
-Spin clockwise for 10 mins
-Spin anticlockwise for 10 mins.
This procedure or routine is what a Function in Sparkster’s platform does, it executes a sequence of commands that you give it. The procedure and device will generate data, letting us know when the delicate wash started. We can calculate the time that the cycle will end, allowing people to know when the washing machine will be available for the next person to use.
The data that would exist here on the blockchain component, where the data stored, which we call a storage cell. Our most recent tech write up gives details of one of our major achievements that we had this week. We were able to get compute nodes to execute a function in five ten thousandth of a second. If you take a second, divide it into ten thousand pieces, in five of those ten thousandth of a second we were able to execute a function on a compute node. In the past we have talked about how the blockchain, where the data is stored, is able to achieve ten million transactions a second.
Fifty two thousand transactions a second have been demonstrated live on YouTube, this is momentous because it shows how we have extremely high performance throughout the entire network whether it's the compute component or the blockchain, where, in most alternative environments there is a bottleneck.
Bitcoin’s TPS limit is three transactions a second, Ethereum's TPS limit is about around seven to thirteen transactions per second. We are able to show you that we can achieve thousands and thousands of transactions a second and also we are able to achieve extremely high performance with the compute nodes.
We are streaming transactions to multiple cells (multiple blockchains) in parallel and that's how we're able to achieve millions of transactions per second per dApp that runs on the network.
How do you make money from running Sparkster’s nodes?
We use probability to assign transactions to different compute and storage nodes on our network. The probability of being assigned a transaction is a determined by the number of tokens you stake. This means we don't have a set threshold, for example, you don't have to use 5,000 SPRK tokens to spin up a node, you can have any number of tokens. This can be from as little as one token. This one token at some point in time, is likely to get a transaction to execute, because it's a probability function. The probability with a small number of tokens is less and with a higher number of tokens is more, but it doesn't mean that the person with the highest number of tokens gets all the transactions.
The mathematics gets more sophisticated because we are looking at things like response time and availability, especially if you are a storage node. It is not just the number tokens, we take in to account but also:
What's your uptime?
What’s your response time in a specific region?
A node that comes and goes is going to have a lower likelihood of being assigned a transaction, compared to one that is always available. The response time works in the same way: if you have a higher response time then you are less likely to get a transaction. This allows us to assign transactions to nodes that are closest to you (by virtue of the way the algorithm works), this means they will have a lower response time responding to your request and therefore your data is closer to you and the whole system works faster.
Although this is a probability function, you are likely to be assigned a transaction no matter how much you contribute. There is essentially an exchange within Sparkster’s environment that's matching what the customer is willing to pay and what the provider of the node is willing to accept in exchange for the use of their device.
Let's say that a company like Panasonic wants to run software on Sparkster’s network and they are willing to pay one SPRK token per function execution, others might want to pay half a SPRK token per function execution. The nodes price that is above the threshold will get excluded from this probability determination. What this means is is that because you are able to decide what price to sell your node for or what price to charge per function execution, this means as more nodes join the network, more people compete within one another to reduce the price that they are charging and that is fundamentally how we deliver on the low price premise.
At the start of this discussion, I talked about how the more nodes to join the network, the price per operation or the price per execution or the price per node falls. This is because the probability distribution is not just a function of how many tokens you stake, but also a function of matching how much the the company like Panasonic is willing to pay. Also who is providing their devices at a price that's less or equal to what other companies are willing to pay. This is how we are able to drive down costs, this is important because we are fundamentally disrupting the price of traditional cloud computing. You are going to see a huge amount of adoption with our decentralized cloud because ultimately it's the cheapest way to run software, It is not centralized and therefore reliable,
But how can community devices be cheaper than traditional cloud computing? Well that’s the discussion of next week’s Tech Tuesday. But as a sneak peak, centralized clouds are run in big datacenters that require large air conditioning devices, diesel generators etc. By using your spare computing power, you don’t incur any of those costs, and therefore your “marginal cost” is near zero.
We will continue exploring this idea of how Sparkster will be delivering this service and understanding some of the concepts that we are building into our network. See you next week!