Hu Yoshida

An All Flash Infrastructure is more than an All Flash Array

Blog Post created by Hu Yoshida Employee on Jul 12, 2016

Earlier this year I posted how we have reached the tipping point for flash, where large capacity flash modules (6.4 TB) from Hitachi, are lower in cost than high performance 15K and 10K hard disk on a relative bit price, when one factors in the 5 year cost of media/tray, maintenance, power/cooling, and floor space.

FMD TCO.png

Gartner published a survey in January which showed a larger list of TCO savings that users experienced with their implementation of flash.

In addition to power, cooling, floor space, and maintenance, the survey responders also listed savings in administration and software licensing which are quite significant. Administrative savings come from the reduction the number of performance issues and licensing savings come from the need to run less instances of applications or data bases or using less processor cores and memory. With these types of savings, it makes little sense to continue to buy HDDs. The future is definitely flash until other non-volatile memories are developed.

Gartner Flash TCO .jpg

AFI versus AFA

This survey did not distinguish what types of flash systems were being used and what the base for comparison was. I suspect it was a survey of All Flash Array users, since that is how Gartner has defined the Flash Market. But looking just at the All Flash Array is short sighted. One really needs to look at the All Flash Infrastructure to get a view of the total cost savings and benefits that one can achieve with flash technology.

 

An all flash infrastructure, is the integration of flash arrays with the rest of the infrastructure and systems management to provide high availability, Integrated management, security, QoS, and scalability.

 

Infrastructure availability

High availability cannot be sustained if the AFA requires off-line upgrades and maintenance, and lacks the ability to do non-disruptive upgrades, active/active clustering, synchronous and asynchronous replication, or does not commit to a 100% data availability guarantee. The Hitachi VSP arrays have all these enterprise capabilities whether you choose the hybrid G series arrays or the all flash F series arrays.

 

Integrated management

Stand-alone management tools may appear to be simple at first, but can add complexity if it does not integrate with current management workflows and automation tools.  Does it support APIs to VMware, Open Stack, Microsoft? Can it be integrated into the orchestration layer of a converged solution? Does it support virtualization that integrates management across a heterogeneous vendor pool of storage. There is a huge pool of legacy HDDs that will be around for a long time that will need to be integrated with pools of flash.

 

Security

Infrastructure security requires that flash arrays provide encryption of data at rest and key management. Of equal importance is the ability to shred the previous data on the flash device and report the results of the shredding for auditing purposes. This is impossible to do on a flash device unless you have software in the flash media that can see flash cells that are masked. This is possible with the Hitachi Flash Media Device.

 

Quality of Service

In an Infrastructure system you have many different workloads and servers competing for flash services and each application expects a certain level of service. When flash devices are new and have a lot of spare cells, performance is great, but that degrades over time as housekeeping starts to interfere with performance, unless you have a flash device like Hitachi’s Flash Module Device (FMD) which has a quad core processor with multi-pathing that takes the house keeping out of the access path. Hitachi also provides an IT Operations analyzer that monitors and provides a view of heterogeneous physical and virtual server stacks, applications, network and storage devices on a single screen for root cause analysis, change monitoring, logical path infrastructure views, and scheduling monitor.

 

Scalability

Typical AFAs are appliances with limited connectivity. Multiple AFAs are required for larger capacity flash data centers which creates silos of information that limit management and insight. Hitachi G and F series storage arrays are built for scalability.

 

Summary

An all flash infrastructure can deliver increased business agility and faster time to market, but It requires the consolidation of business operations on a single flash platform provided by a proven enterprise storage infrastructure vendor who has the R&D resources to integrate future enhancements in non-volatile memory systems.

Outcomes