Google dives into elusive cloud file storage market
Google Cloud Filestore may be the last piece of the puzzle for the vendor's storage services, but like its competitors' file offerings, it's far from a finished product.
Cloud file storage remains a tough nut to crack for the hyperscale providers, but it's a key ingredient to their ongoing success with enterprise customers.
Google Cloud Filestore, available in beta in July, is a managed service for enterprises to store unstructured data in a shared file system, with network-attached storage that integrates directly with Google's cloud VMs and Kubernetes instances.
These services are important for enterprises whose applications depend upon local file servers and, if there isn't an equivalent on the public cloud, it's enough reason to stay put. As more data moves to the public cloud, the market has shrunk for other types of on-premises storage, such as home directories, but that's not the case with file storage, one expert said.
"In fact, with the addition of emerging workloads for AI, machine learning and industrial IoT, it's actually expanding," said Julia Palmer, a Gartner analyst.
With this initial release, Google will tackle high-performance storage for applications that require high throughput, low latency and high IOPS. File storage is often associated with legacy systems, but, in Filestore, Google highlights content management systems, website hosting, image rendering and virtual workspaces for artists.
There are two tiers, and customers are charged for provisioned capacity, with a maximum size of 64 TB. Premium storage is designed for 700 MB per second and 30,000 IOPS, and costs $0.30 per GB per month. The lower tier costs $0.20 per GB per month, but performance scales with capacity and peaks at 10 TB and above.
Jellyfish Online Marketing, a marketing agency in Reigate, England, runs campaigns and the applications and websites behind those efforts. Jellyfish wanted to use Google Kubernetes Engine (GKE) with open source tools such as WordPress, but the Google Cloud storage system was insufficient for an architecture that runs multiple copies of WordPress in parallel. Previously the company either built its own centralized cloud file storage system, which created significant overhead, or relied on other cloud providers.
Filestore is the "missing piece in the puzzle" -- a native storage integration that lets Jellyfish fully embrace Google for its managed Kubernetes clusters, said Ashley Maloney, lead DevOps engineer at Jellyfish.
"It's a big deal for us to be able to natively use GKE without a bunch of custom modules outside of a traditional open source stack," he said. "Having Filestore there allowed us to have that native integration with Kubernetes while giving us the benefits of scalability and high uptime."
Cloud file storage still means tradeoffs for enterprises, providers
Jellyfish's use case was relatively simple but effective, as the native integration was more important than the breadth of file storage features. But enterprise clients with legacy systems will expect more before they ditch their existing setup for cloud file storage.
It's difficult to build file storage systems, and even more so while maintaining reliability and performance in distributed systems, which is likely why Google added Filestore last, Palmer said. File storage is more performance-intensive than blob or object storage, with higher IOPS and throughput demands. There are also far more features compared to cloud-native storage, which has fewer protocols and is well-suited to the infinite scale of public cloud.
File storage is going to need to get better in order to improve enterprises' ability to really utilize public cloud.
Julia Palmeranalyst, Gartner
Typically file systems are written to be tuned specifically for performance or throughput. Google appears to have a robust roadmap for Filestore, and it makes sense to go after performance first because it opens Google Cloud to more companies that want to use its analytics tools, Palmer said. But that route likely comes with tradeoffs in scale, protocols or data services such as snapshots.
Google lags behind competitors on its cloud file storage options, but nobody has cornered the market. AWS added Amazon Elastic File System (EFS) in 2016 after a lengthy testing period, but it still lacks some of the features companies require. For example, as is the case with Filestore, EFS doesn't support server message block (SMB), which is the standard Windows protocol for file storage. Microsoft has its own service called Azure Files, and earlier this year it acquired Avere Systems, which also offers hybrid cloud file storage, though it isn't expected to be fully integrated with Azure until later this year.
"They all have their shortcomings and all three are working to close them," Palmer said. "They all agree that file storage is going to need to get better in order to improve enterprises' ability to really utilize public cloud."
