Enterprises now have a new option for running memory-intensive applications on Google's cloud platform.
The company this week announced beta availability of n1-ultramem, a new family of memory-optimized virtual machine (VM) instances that Google says is well suited for applications such as data analytics, enterprise resource planning and genomics.
The new machine types offer more memory and computing resources than any other virtual machine type that Google currently offers and is ideal for resource-hungry High Performance Computing apps as well, according to the company.
"These VMs are a cost-effective option for memory-intensive workloads, and provide you with the lowest $/GB of any Compute Engine machine type," Google product manager Hanan Youssef, wrote in a blog May 15.
The n1-ultramem machine types provide enterprises running workloads in Google's cloud with new price-performance configurations. "Now, you can provision compute capacity that fits your exact hardware and budget requirements, while paying only for the resources you use," said Youssef. With the new machine types, Google now offers VM configurations starting with 1 vCPU all the way to machine types 160 CPUs and multiple terabytes of memory.
Each of the new machine types is powered by 4 of Intel's Broadwell processors (Xeon Processor E7-8880 v4) and come in three pre-defined configurations. The lowest-end configuration comes with 40 virtual CPUs and supports 961 GB of memory. The mid-tier system supports double the number of vCPUs and 1.9 terabytes of memory while the highest-end system supports 160 vCPUs and 3.84 terabytes of memory.
Google is positioning the n1-ultramem-160 CPU machine type as big enough and ideal for running's SAP's HANA in-memory database. With the new machine types, enterprises no longer have to keep their database on-premise while apps move to the cloud. The VMs are large enough to allow enterprises to run both the application and the SAP HANA database on Google's cloud platform, Youssef said.
Pricing for the new machine types starts at just over $3,220 per month for the lowest capacity configuration with 40 vCPUs and 938 GB of memory. Enterprises that sign up for the highest capacity system with 160 virtual CPUs and 3.8 TB of memory will pay $12,885 per month.
As with its other machine types, Google's n1-ultramem machine types are also available under the company's so-called pre-emptible VM pricing model. With pre-emptible pricing, enterprises can run their workloads on Google's cloud infrastructure at a substantially lower rate than usual. But the systems can be pre-empted when Google requires them to run other workloads any time it requires additional compute capacity. Google has described the option as ideal for large batch processing jobs.
Enterprises that sign up for the pre-emptible model can get the lowest capacity n1-ultramem configuration at just over $970 per month compared to the regular $3,221. Similarly, pre-emptible pricing for the 160-vCPU-configuration is $3886 monthly compared to $12, 885 for the regular configuration.
Google's n1-ultramem machine types are currently available for beta evaluations to Google's cloud customers in certain U.S. and European regions and will become available elsewhere soon.