Data-driven hyperscale companies serve so much data, so quickly, that they are building their own servers. Unlike traditional enterprises running classic applications in four to five year cycles, hyperscale companies upgrade their servers and storage more frequently—often every two to three years. Their lightning fast service evolution mirrors the pace of the underlying infrastructure. Everything is accelerated.
Hyperscale computing, used in cloud and big data environments, encompasses the infrastructure and provisioning to effectively scale from several to thousands of servers. Hyperscale companies have unique needs compared to enterprise organizations. We've designed Fusion ioScale with the input of existing hyperscale market leaders to maximize the simplicity of the all-flash datacenter and meet unique hyperscale needs.
When aiming for application transactions, flash is far more efficient than the humble disk drive in scaling performance. But flash pricing has caused some companies to reserve flash memory for just their most critical applications and databases. Echoing this, IDC recently issued a report on enterprise flash adoption rates, noting “pricing [of flash] remains a key hurdle for broad adoption. Early market dynamics suggest that … as solutions become more affordable, IDC expects to see increased adoption of solid state solutions.”
Breaking Through Price Barriers
With the introduction of hyperscale flash memory at $3.89/GB in volumes as low as 100 units, we enter the era of the all-flash hyperscale datacenter. It is now affordable to accelerate an entire application server tier with flash. By ditching disk, companies can offer speedy user experiences with consistent quality of service, paving the way for new web applications.
Servicing more users who are accessing more data and collaborating on real time can only be accomplished with flash-based solutions. If you have hit a wall for how long you can wait for your disk-based application, server-based flash memory breaks those limitations. Hyperscale companies can also take advantage of native flash memory APIs to guarantee they get the most value out of their infrastructure.
Disk may still hang around in archiving and cold storage uses, but with flash more accessible on a large scale, disk is likely to recede. Disk is slow, small in capacity – especially when short stroking – takes more power to run, and generates heat that in turn drives cooling costs. With affordable flash now on the market for webscale companies, the all-flash hyperscale datacenter is here to stay.
Check out more about ioScale here.