Storage giant EMC made a flurry of announcements this week. Here are its expectations for what businesses will do next in the world of storage.
In the face of a global spending slowdown on high-end storage, fierce price pressure from public cloud services and demand for cheap commodity storage for scale-out clusters, US storage giant EMC is recasting its premium VMax family as more than just a storage array, building closer ties with public cloud services and moving into the commodity storage business.
Remodelling the high-end
High-end storage has been the "stalwart of the EMC portfolio for the last 20 years", according to Jeremy Burton, president of products and marketing at EMC. The latest incarnation of EMC's flagship array, its VMax3 series, attempts to turn it into a platform for consolidating datacentre infrastructure.
As well as VMax3 being a storage array, EMC wants the boxes to run the data services that are often handled by separate boxes. These services could include acting as file gateways, fully automated storage tiering, EMC Vplex replication and some ETL (Extract Transform Load) functions.
To support these additional tasks, the VMax3 boxes unveiled on Tuesday have various hardware and software improvements. The hardware in the new 100K, 200K and 400K Vmax appliances is composed of what EMC calls engines. The family scales up to maximum of eight engines for a total of 16TB cache, 384 cores and 5760 drives - a mix of HDD and SSDs, all of which can deliver 6.3 million IOPS and network bandwidth of 1792 Gbps.
100K: Up to two dual-controller engines. 6-core 2.1 GHz Intel Xeon E5-2620-v2 processor. 24-cores per engine to a maximum of 48 cores. Maximum drives per system 1440 / usable capacity 496TB.
200K: Up to four dual-controller engines. 8-core 2.6 GHz Intel Xeon E5-2650 v2 processor. 32-cores per engine to a maximum of 128 cores. Maximum drives per system 2880 / usable capacity 2.04PB.
400K: Up to eight dual-controller engines. 12-core 2.7 GHz Intel Xeon E5-2687-v2 processor. 48 cores per engine to a maximum of 384 cores. Maximum drives per system 5760 / usable capacity of 3.97PB.
Individually the specs of the new VMax appliances are:
Each engine has dual 56 Gbps InfiniBand interconnect
The boxes pack disks more densely than previous VMax releases - with the new Viking disk enclosures able to contain 120 1.6TB solid state drives in 3 rack units, for a total of 64TB of storage per rack unit or 640 TB per floor tile. The 100K can have two drive array enclosures, the 200K four and the 400K eight.
VMax3 arrays storage is composed of about 90 percent hard drives and about 10 percent solid-state drives.
To help these new boxes exploit this additional hardware EMC has tweaked the VMax's software. The appliance's "dynamic virtual matrix" helps ensure the available processing power and memory is allocated in such a way to deliver optimal and predictable performance for workloads. Allowing the VMax to run a greater range of EMC data storage services is the new Hypermax OS, a converged storage hypervisor and operating system. The top of the range VMax 400K can support about 40,000 virtual machines.
"We can now run not just data services but infrastructure applications themselves, things like replication and backup and management consoles," said Burton.
EMC claims VMAX3 offers up to three times faster performance and 50 percent lower total cost of ownership than previous generation VMAX systems, accelerating Oracle, SQL, SAP transactional processing, high bandwidth data analytics and file workloads. Overall VMax3 delivers storage at a cost of about $3 per GB and $1.85 per IOP, according to EMC, which also promises six nines availability for the appliances.
Another major focus for EMC is making its high-end storage easier to manage, with Burton admitting its premium Symmetrix range had traditionally not been thought of as "agile".
To make configuration easier EMC has introduced VMax Advanced FAST Suite. The suite allows storage admins to set what EMC calls service-level obligations (SLOs) for certain workloads, such as the maximum latency for accessing data, and then allows the software to decide where data should be placed to meet those targets.
In a demo a user was shown choosing from bronze to diamond service levels, which offered from 40ms to sub-1ms latency.
The service is designed to save storage admins from having to spend time manually tiering data. Burton said EMC wanted to make VMax "as easy to manage as a public cloud service".
Another way that EMC plans to reduce infrastructure and network load is via VMax's ProtectPoint backup service. ProtectPoint introduces the ability to backup data direct from VMax to a Data Domain system, without that data having to be passed through intermediate back-up servers, or boxes handling services such as deduplication.
Roger Cox, server and storage specialist with analyst house Gartner, said the move to make VMax easier to manage is a significant step forward.
"The most important thing they've done with this announcement is to substantially simplify the way you go about provisioning and managing these systems."
Cox also said ProtectPoint was further evidence of EMC "positioning VMax as more than just a storage array and more and more as a datacentre services platform".
The VMax3 series appears to be at least comparably, if not better specced, than the latest Hitachi Virtual Storage Platform model, the G1000, he said.
VMax3 and the Hypermax OS are expected to be available in the third quarter of this year. Embedded file and ProtectPoint will be available in the fourth quarter for additional license fees.
Working with, and against, the cloud
EMC is taking various steps to both compete with and benefit from the rise of public cloud services - whose low-cost storage offerings threaten to cannibalise some of its business.
On the one hand EMC is taking steps to make it easier to push data from its storage to public cloud services, acquiring Twin Strata, a company that makes the cloud storage gateway technology CloudArray.
CloudArray is a caching appliance that can automatically replicate the data stored on it and move it cloud storage. Twin Strata co-founder Nicos Vekiarides said the company plans to integrate CloudArray into the VMax3 data services platform "to allow users to automatically tier workloads even more seamlessly for off-premise storage capacity expansion, data protection and disaster recovery". CloudArray supports various public cloud storage services, including Amazon S3.
EMC is also making it easier to use third party management tools to provision and control virtual pools of storage running on its VMax arrays. Version 2.0 of EMC's ViPR software defined storage layer presents what EMC describes as a "consistent interface" for managing these storage pools to VMware and Microsoft management layers.
EMC's other approach to competing with public cloud storage pricing is to move into selling commodity storage appliances.
As well as managing storage on EMC arrays the ViPR controller, and version 3.5 of EMC's storage resource management suite, can automate the provisioning and management of storage pools based on non-EMC, commodity hardware.
EMC is now building appliances using "commodity hardware" that use ViPR data services to provision and manage storage pools. These appliances, badged as Elastic Cloud Storage (ECS), are designed to offer large pools of low-cost storage at a price that competes with cloud storage offerings from the likes of Amazon and Google.
EMC CEO David Goulden predicts that while demand for commodity systems like ECS will see huge growth, companies will still continue to buy high-end appliances like Vmax for a long time to come.
"We're not talking about anything that's going to transition very quickly but the growth rate on ECS-type systems is going to be exponential," he said, with that rapid growth coming from capturing and analysing far more data than in the past.
"There's a dollar opportunity on VMax-type systems because again datacentres are growing by 25 to 30 percent [each year], so that's also a market that has potential."
The first 3PB ECS system will be used by the Vatican library to digitise their archives and make them available via the web.
This week EMC also released two new Isilon network attached storage products, capable of roughly twice the performance of current models, as well as refreshing Isilon's OneFS operating system to improve performance and add support for new data access protocols. The changes are designed to widen the workloads Isilon is suited to, making it possible to use the scale-out clusters for big data analytics, cloud applications and mobile sync and share.
The Isilon S210 is an upgrade of the S200. It is available with 7.2TB, 14.4TB, 21.6TB and 28.8 TB nodes, and can be scaled up to 144 nodes for a maximum cluster of 4.15PB. The S210 is capable of 3.75 million IOPs per cluster, is optimised for high transaction workloads and can deliver twice the performance of its predecessor, according to EMC.
The other new offering is the X410, an update of the X400 series. The X410 offers 70 percent increase in throughput - which scales to more than 200 Gbps on a 5.2 PB, 144-node cluster. The product is aimed at supporting Hadoop analytics, high performance computing and enterprise file applications.
The latest version of the OneFS operating system, version 7.1.1, provides support for up to 1PB of flash-based cache to speed data retrieval. OneFS 7.1.1, the S210 and X410 will be available this month.
EMC also plans to broaden the protocol and access methods for data stored on Isilon clusters - adding support for Hadoop Distributed File System 2.3 and OpenStack SWIFT native Object by the end of this year.
The firm is also teaming up with Pivotal, the joint venture it runs with Vmware, to offer what it calls a Data Lake Hadoop Bundle, which is based on Isilon NAS and designed for storing and analysing very large unstructured data sets.
More than 400 organisations are using the big data processing framework Hadoop on Isilon, according to Burton.