Photobank kiev - Fotolia
When should you move to the all-flash datacentre? When you’ve got 80TB to 90TB of mixed workload data that can be stored on it.
That’s the view of IDC analyst Eric Burgener, who recently set out this view in a white paper.
In the document – sponsored by flash array supplier Violin Memory – he says the cost per GB for flash when data reduction is taken into account is now down to around $1.25.
Although this still has some way to go in terms of raw cost, with that of hard disk drive (HDD) at $0.70 or $0.80 per GB, the secondary economic benefits – such as savings in numbers of servers and power and cooling costs – make all-flash a viable proposition for 80/90TB-plus mixed workloads.
“Everyone we have spoken to says they would deploy flash if it cost the same as HDD. Cost is the main obstacle. $/GB raw is not a relevant metric, but $/GB effective is, especially when you consider the effects of data reduction and ratios on average of 5:1,” said Burgener.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
“People need to look at the secondary economic benefits of flash when calculating TCO [total cost of ownership]. And if you can do mixed workload operations to 80TB or 90TB, you can probably achieve significantly better TCO than with HDD.”
According to Burgener, the key is that flash drives offer massive advantages over spinning disk in terms of input/output (I/O), density to achieve input/output operations per second (IOPS) and power and cooling.
Flash array makers are also increasingly building in the kind of features enterprises require, such as, replication for disaster recovery, security, snapshots and clones. “The kind of things you find on a VMAX and VNX, for example,” said Burgener.
“As you consolidate workloads, it improves the advantages of the performance density of flash deployments,” he said.
Burgener cited IOPS figures for a standard 15,000 rpm HDD as 180-200 IOPS and for flash drives as 10,000 to 20,000 IOPS.
“So if you’re trying to hit 500,000 IOPS, you need 2,500 200 IOPS drives or 25 20,000 IOPS ones,” he said.
“Flash uses roughly half the energy of a HDD and the speed of flash allows Intel CPUs [central processing units] to be fed without them waiting around for data – which was always a problem with HDDs – and so achieves good utilisation and means fewer servers are required.”
“With those considerations in mind, 80/90TB is the point at which you start to make savings of $150,000 on power and cooling utilisation, for example, on a four year TCO on medium/large systems that cost $300,000 to $400,000.”
IDC forecasts that more than 60% of primary storage spend will be on flash by 2019.
Read more about flash storage
- In this special report, Computer Weekly looks at the product offerings and main trends in the SSD all-flash array market.
- PCIe SSD is a new way of adding the speed of a solid-state drive (SSD) to server and storage devices. What are the benefits, drawbacks and adoption rate of PCIe SSD?