Why Ethereum developers are buzzing about The Graph — even though it's still centralized

Ethereum-based virtual world Decentraland allows users to create, trade and own assets in a digital reality. Through one of the project’s core features, Marketplace, users can transfer and sell land, purchase digital items and monitor other users’ assets.

Every displayed product in Marketplace is associated with certain pieces of data that are scattered around different transactions on the Ethereum blockchain. Because of this, it isn’t an easy task to fetch and consolidate all that information into the single data point that a user sees. That’s why Decentraland turned to The Graph.

The Graph, a San Francisco-based startup, has developed an indexing protocol that organizes all the information on the blockchain in a more efficient way and helps developers of Ethereum-based applications run their frontend operations. The Graph’s tool can be used to find and retrieve data from Ethereum, similar to the way other indexing tools can be used to do the same for traditional, centralized databases.

Decentraland is just one of many fairly well-known Ethereum projects now using The Graph’s technology to index their data and present it in a user-friendly way. The roster includes the video streaming service LivePeer, the Ethereum Name Service, decentralized development fund MolochDAO, and market-making platforms Uniswap and Compound, among others.

Indeed, according to The Graph, it has seen query volume grow 50% per month since the launch of its hosted service in January 2019. And last month, the firm said it processed over a billion queries for the first time. Investors have taken notice, too. Late last month, the San Francisco-based company announced that it raised $5 million via a token sale that involved several crypto funds including Digital Currency Group and Coinbase Ventures.

The Graph has clearly arrived. There’s still one issue, though: it’s centralized. Users must rely on the startup’s own Google Cloud database, and if it goes down — like it did last month — the service will be unavailable.

Now, the firm says it will use the new funding to achieve something that has flummoxed many blockchain projects that came before it before it: decentralization.

Subgraphs

Much of The Graph’s recent growth has been driven by so-called subgraphs. Subgraphs are open APIs that applications can query using GraphQL, a language for APIs that can be deployed similar to the way SQL is used to to query and retrieve data in a traditional relational database. Developers build subgraphs by defining specific data sources, supplying a method for transforming and processing that data, and making it possible for the data to be queried. Once the subgraph has been built, anybody can query it.

DApp developers can find new subgraphs using Graph Explorer, through which they can browse and experiment with APIs using The Graph’s built-in GraphQL Playground feature. They can replicate the information they receive inside their applications, avoiding the need for backend servers.

For instance, video streaming platform LivePeer has developed a subgraph hosted by The Graph to make it possible for its developers to access a massive amount of data quickly and give users visibility on what’s going on in the network.

LivePeer founder and CEO Doug Petkanics says that has helped decrease LivePeer’s load times, from up to 30 seconds to load a page down to under a second. It has also added functionality to the platform, allowing for the display of features and data that was not possible without an indexing protocol, says Petkanics.

Put simply, the Graph’s technology makes it vastly easier for developers with limited knowledge about Ethereum and blockchains to build web apps in a more traditional way, says Esteban Ordano, an advisor for Decentraland, which has also developed a subgraph.

“We’ve been using horses our whole life and trying to build something with wheels,” Ordano says. “Then The Graph came along and showed us the car.”

A single point of failure

For all the buzz around The Graph, however, it is still a centralized service. And on the morning of June 24, The Graph’s users got a harsh reminder of that.

That day, data requests began to fail, apparently because volumes got too high. “Despite having 50% headroom in our Google Cloud database, on June 23 our database CPU maxed out at 100%, causing requests to fail,” co-founder Yaniv Tal explained in a post-mortem blog post. The disruptions affected the frontend operations of several applications that were relying on it.

Although LivePeer didn’t fully collapse during the outage, its frontend application, which LivePeer developers were using to view what was going on in the network, was affected. Users could still access the data but would have had to run their own full Ethereum node to do so.

Nonetheless, the episode exposed the risk of relying on a centralized tool for such an important function. “When The Graph’s servers go down, they can take down all the apps built on top of it,” says Petkanics. “That’s a pretty risky position to be in, to build your business on.”

The outage also caused Decentraland’s Marketplace to go down for a few hours, says Ordano. “We’re looking forward to the launch of the decentralized network because we won’t have to rely on a single node of theirs.”

Tal is in agreement. “Yesterday’s incident highlights the importance of decentralization at every layer of the stack,” he wrote in the post-mortem. “We don’t want teams to have to trust us.”

Then, in a blog post published June 30 announcing the new fundraise, Tal promised to transition the indexing tool from a hosted service to decentralized network by the end of the year. Once the transition is complete, developers will no longer have to publish their subgraphs directly to The Graph’s hosted service. Instead, they will be able to rely on “independent indexers who run nodes and process queries in an open marketplace,” he wrote. “Curators will be able to join the network and organize data and signal which subgraphs are useful and accurate.”

“Developers will need to start paying for fees in an open marketplace where indexers compete to provide the best indexing service at the lowest price,” Tal explained. “The introduction of these new roles will ensure that decentralized applications are built on a serverless foundation with no point of failure.”