Google has put persistent high-IOPS solid state disk (SSD) storage into open beta for Compute Engine users. That means Google Cloud Platform users can access SSDs for workloads, which provides faster I/O and less redundancy than traditional magnetic media -- and comes in handy for databases and analytics.
With high-IOPS SSDs, Google claims that "performance scales linearly from one to four partitions, with the full four partitions executing up to 680,000 random 4K read IOPS and 360,000 random 4K write IOPS. That's eight times the write IOPS/GB and 15 times the read IOPS/GB than with [Google's] SSD Persistent Disk [service]."
Google is charging 21.8 cents per gigabyte per month, which translates to 0.03 cents per gigabyte per hour. That’s not bad! But how do you know you need that level of performance? If you're using big data systems running Hadoop or Hadoop-like technology, high-speed data analytics, or data acquisition systems -- anything that goes to disk a great deal -- you're a candidate for the high-IOPS SSD service.
However, this doesn't mean everything has to use cloud-delivered SSD. Indeed, only specific workloads benefit from this technology, so you must define the workloads and perhaps do some benchmarking to understand the true benefits. Your objective should always be to maximize the price/performance ratio; this technology simply provides you with more options.
Still, you can expect high-IOPS SSD to quickly become common across cloud providers. The price of SSD memory is dropping quickly, and most enterprises are becoming addicted to the price/performance ratio that SSDs bring. I can even see a day when spinning-disk technology will be outmoded.
Of course, the infrastructure investments to add high-IOPS SSDs could translate into higher prices. But for now, the price of public cloud services continues to drop.