Containers Not Dealing a Death Blow to Public Clouds Anytime Soon
With all the talk of containers recently you'd be forgiven for thinking that public clouds filled with old-fashioned virtual machines were a thing of the past.
Not so. In fact, public cloud providers such as Amazon, Google and Microsoft are all going strong and have been busy coming out with some interesting new offerings.
Microsoft has been busy thinking up ways to make it easier to run compute and storage-intensive applications — like SQL Server and MySQL, NoSQL databases like MongoDB, and even data warehouses — at top speed.
Its solution is a variant of the G-series monster VMs that the company is calling the GS Series, and which essentially is a line of VMs that have the computing power of the G-Series VMs combined with the storage performance of Premium Storage.
The G-Series VMs range from 2 core to 32 core machines, making the largest ones very grunty indeed. Now combine that with up to 64TB of storage, 80,000 IOPs and 2,000MB/s of storage throughput, and you have a very capable GS-series machine.
In fact, the GS-series offers more than double the disk throughput of any other VM offered by any other hyperscale public cloud provider, according to Microsoft. The GS (and the G) series also offers more than double the network bandwidth (20Gbps) of the closest VMs of any other similar provider — again, according to Microsoft.
While upping the ante at the top end, Microsoft is also cutting prices further down the VM food chain. The company has announced it is reducing the prices by as much as 27% as of October 1st on its more modestly specced D-series and DS-series VMs — which are designed for the most common workloads.
Meanwhile, Google has announced it is launching a new high-end series of 32-core VMs designed for large-scale compute and storage-intensive work such as graphics rendering, video transcoding, and large MySQL and Postgres instances.
The 32-core VMs don't have a catchy name, but they do come in three flavors:
- Standard: 32 virtual CPUs and 120 GB of memory
- High-memory: 32 virtual CPUs and 208 GB of memory
- High-CPU: 32 virtual CPUs and 28.8 GB of memory
Google Compute Engine Autoscaler Makes Its Debut
Google's also been hard at work on an auto-scaling system, and has finally announced the general availability of its new Google Compute Engine Autoscaler.
So what exactly is it? The answer is it's a system that dynamically scales instances in response to load conditions. You define the ideal utilization of your group of compute instances, and Autoscaler will add instances when needed and remove them when traffic is low, Google says. The benefit of this is it reduces costs as you don’t have to buy and hold spare capacity. In fact, Google promises it will allow you to forget about capacity planning and load traffic monitoring all together. Not bad, eh?
So what's Amazon been up to in its AWS cloud? Nothing as dramatic as new VMs, but it has introduced a brand new storage class for S3: Amazon Standard Infrequent Access (Standard - IA). Standard IA sits between Amazon's Standard storage and its ultra-low cost but high-latency Glacier storage, offering instant accessibility at low cost, with slightly lower availability than Standard (99% instead of 99.9%). Sounds familiar? That's because it sounds eerily like Google's Nearline storage that sits beneath Google's Standard service.
Amazon Standard IA is also a good deal cheaper, at $0.0125 per GB for the first TB per month, than Amazon's Reduced Redundancy storage, which costs $0.024 per GB for the first TB per month. And to ensure it remains competitive, Glacier prices have been slashed by up to 36% in US East, US West and Europe (Ireland) to just $0.007 per GB per month. (When the service was launched in 2012, Glacier pricing was $0.01 per Gb per month.)
So there you have it. Bigger VMs aimed at demanding workloads, cheaper standard VMs, and cheaper storage. There's plenty of life left in the container-free public clouds.
Paul Rubens is a technology journalist and contributor to ServerWatch, EnterpriseNetworkingPlanet and EnterpriseMobileToday. He has also covered technology for international newspapers and magazines including The Economist and The Financial Times since 1991.
Read more on "Cloud Computing" »