Google's local SSD is now generally available on its cloud platform, offering a competitive price point and quality...
to potentially match Amazon's SSD alternative.
The feature went into beta in October and is now generally available to Google Compute Engine customers in all regions, though not currently in every zone within those regions. Customers can attach between one and four SSD partitions of 375 GB to a virtual machine (VM) for dedicated use of the partitions.
SSD is almost a required feature, and Google cloud customer Workiva, a financial reporting software developer, plans to look at local SSD as it does more caching, said Dave Tucker, senior director of platform development for the Ames, Iowa-based company.
Workiva began as a Google App Engine customer but is relying more on Google Compute Engine as it expands its set of features, Tucker said.
"I'm encouraged by progress in these services," Tucker said. "I'm sure it's going to be a valuable piece in the puzzle."
Google's local SSD provides 680k random 4K read IOPS, but, unlike persistent disks, it doesn't provide redundancy. The feature is good for high-demanding applications that provide their own replication and as a supplement for memory due to its high IOPS and low price per GB, according to Google, which already offered persistent disk SSD.
Aerospike, Inc., based in Mountain View, Calif., used the local SSD and attained RAM-like performance and 15X cost advantage as a supplement for RAM on their NoSQL database.
Google local SSD versus AWS storage
Google trails some of the other cloud vendors in terms of market share, but the quality of this feature and others shows the company clearly is investing in the platform -- and customers of other vendors are becoming more interested, said Brian Bulkowski, Aerospike founder and CTO.
"This offering is at the highest level of performance -- similar to Amazon and ahead of some of the other cloud vendors," Bulkowski said.
There isn't much separation between Amazon Web Services (AWS) and Google's SSD options, Bulkowski said. Google customers can restart a full instance and keep the data because the local SSD is more of a persistent tier than Amazon's ephemeral disk. However, AWS customers have a wider variety of instance types to choose from, he added. AWS' latest Elastic Block Store (EBS) volumes can be auto-recovered.
Local SSD costs $0.218 per GB per month. For customers purchasing local SSD attached to VMs, the price is $0.0003 per GB per hour, while price per-IOPS comes to $0.00048 per read-IOPS per month, Google said.
Amazon charges $0.10 per GB per month for its EBS General Purpose SSD and $0.125 per GB per month for its EBS provisional IOPS SSD, in addition to $0.065 per provisioned IOPS.
Google's pricing is competitive and, unlike many other cloud providers, Google separates SSD drives from compute capacity, said Owen Rogers, senior analyst for 451 Research LLC, based in New York. This gives customers the freedom to choose how much capacity they want, as well as how much SSD storage they require, without being constrained by fixed sizes, he added.
"This is good news for end users, as it means they can improve cost-efficiency and performance by choosing the best combination of resources for specific workloads and applications, without being forced into buying too much or too little capacity due to the fixed size of virtual machines," Rogers said.
The price of a typical application running on Google Compute Engine fell 7% from October to November 2014 as a result of cuts on bandwidth, SSD and SQL, according to 451's Cloud Price Index.
There are plenty of cloud-based storage options from other providers, but the option of in-memory, direct attach storage could be a sweet spot for Google if it's presented properly and reaches a wide enough audience, said Robert Mahowald, vice president at IDC, a Framingham, Mass.-based market research firm.
"They don't have any choice for redundancy so it's a bit of a challenge," Mahowald said. "But the speed and the cost are attractive."
Uses would all be related to databases, such as de-duplicating massive data files or cataloging block storage that requires a ton of memory churn, Mahowald said. The challenge with this type of feature is configuration and knowing how long something has to exist and when to burst between local and non-local SSD, he added.
"For customers, there's some skill and sophistication involved with thinking through workload scenarios," Mahowald said.
Trevor Jones is the news writer for SearchCloudComputing. You can reach him at email@example.com.
Benchmark testing comparison: AWS vs. Google SSD
Google vs. AWS SSD: Which is the better deal?