The announcement of EVO:RAIL at VMworld and VMworld Europe 2014 showed the importance VMware is putting into hyper-converged infrastructure. The first of these two articles examined what technology EVO:RAIL is based on and in what respect the appliances of the different OEM partners vary. This second article shows what use cases apply to EVO:RAIL and discloses the pros and cons. The original German language version of this article can be found on SearchDataCenter.de.
VMware EVO:RAIL is converged infrastructure developed into hyper-converged infrastructure. Beneath compute, storage and networking resources, EVO:RAIL comes preconfigured with integrated software like firmware, drivers, hypervisor and management tools as a "datacenter-in-a-box."
EVO:RAIL was developed with easy management in mind
According to Mornay Van Der Walt, vice president Emerging Solutions Group SDDC Division at VMware, the highest priority while developing EVO:RAIL was making the deployment process as easy as possible with minimal administration hurdles. As the consumption model of modern IT changes, IT shops are searching for easy options to provide their employees. Based on this trend, Van Der Walt explained during a background talk at VMworld Europe 2014, EVO:RAIL was developed with a big focus on easy deployment.
At its booth at VMworld Europe 2014, Fujitsu showed how the new EVO:RAIL engine enables the deployment of EVO:RAIL within less than 15 minutes, at least under ideal conditions. It also showed the way that the EVO:RAIL engine is capable of automatically detecting new appliances and integrating them without downtime into existing EVO:RAIL deployments.
Due to its easy handling and its ability to scale quickly, VMware positions EVO:RAIL as a rapid scale-out solution for virtualization or VDI, mainly in branch offices without big IT departments. For Val Bercovici, who has worked for NetApp for more than 16 years, this narrow focus discloses one of the most obvious drawbacks of EVO:RAIL: as the trend towards using cloud services continues, the target group for hyper-converged infrastructure (small companies or branch offices with small IT departments) is declining continuously. Even VDI, which is one of the most plausible use case of EVO:RAIL, faces strong competition through technologies like Desktop as a Service (DaaS).
Because of this, Bercovici doesn´t believe EVO:RAIL, or hyper-converged infrastructure in general, will be a huge success. In his opinion NetApp, with its converged FlexPod architecture, is much more flexible and could provide a wider set of workloads. NetApp FlexPod is based on Cisco’s UCS platform and is available in Express, Select and Datacenter editions. According to NetApp, it is capable of hosting not only simple virtualization workloads but even workloads regarding high performance computing or big data scenarios. Apart from that statement NetApp recently announced an EVO:RAIL appliance of its own despite VMware being a child company of EMC, one of the biggest competitors of NetApp.
New EVO:RAIL products could be right around the corner
Tony Parkinson, vice president EMEA Enterprise Solutions at Dell Inc., noted that in the long run there probably won´t be just two sets of EVO:RAIL appliances, especially as they only differ in the software they use. To overcome the inflexibility of hyper-converged architecture, there is a certain need to offer various models with different hardware for different use cases. You simply can´t plug in a better graphics processing unit in EVO:RAIL as you can do with standard servers.
Therefore, EVO:RAIL is just the beginning of VMware's venture into the hardware market, as VMware itself points out EVO:RAIL is the first in a whole family of hyper-converged infrastructure appliances. Parkinson said he could imagine VMware integrating its vSphere Distributed Switch into future EVO:RAIL appliances.
If VMware really is heading in this direction and EVO:RAIL will be available in different versions with varying hardware/software for different use cases, it certainly could satisfy complex workloads as converged infrastructure. At this point the easier management, as well as the better scalability and performance, could indeed be a serious threat to existing vendors of converged infrastructure like NetApp, Cisco and others. That might have been a reason for NetApp to change its mind and announce an EVO:RAIL appliance of its own.
But there are cons of hyper-converged hardware that even EVO:RAIL can´t solve. Because of the tight integration of compute and storage it´s impossible to scale both independently, said Rhadika Krishnan, vice president Solutions and Alliances at Nimble Storage.
For many workloads, the need to scale storage resources is in higher demand than the need to scale compute resources. But with EVO:RAIL, or hyper-converged hardware in general, it´s not possible to scale the one without automatically scaling the other, as the only way to scale storage at some point would be to buy another appliance. In that case you would buy compute resources you don´t need just to satisfy your storage needs.
For Krishnan, the important point with hyper-converged infrastructure is to think about your needs of scaling storage and compute independently. Krishnan expects the hype around EVO:RAIL to flatten in the coming months because the market shows continuous waves of integrated and less integrated products. Nevertheless, Nimble Storage itself offers SmartStack for ROBO, a reference architecture for converged infrastructure on top of Cisco UCS.
Integrated deduplication and backup as unique selling point of SimpliVity
Like the success of virtualization has led to consolidation of physical servers, it is now time to further consolidate hardware through (converged and hyper-converged infrastructure. , As a result of this ongoing consolidation IT management focuses more on operating expenses, not so much on capital expenses any more. According to Johannes Kunz, who is responsible for the channel distribution in the DACH region at SimpliVity, hyper-converged infrastructure like SimpliVity OmniCube or VMware EVO:RAIL can help to drastically reduce OPEX spending.
Wolfgang Huber, Regional Director Central Europe at SimpliVity, says there is one big difference between EVO:RAIL and OmniCube: integrated solutions for deduplication and backup. As Huber further explains, by now these are critical enterprise functions no company would easily ignore. Both Huber and Kunz admit that EVO:RAIL is just be at the beginning and there might be integrated deduplication and backup in the future. But by, now both assure SimpliVity has the most complete offering and provides these functions for quite a while.
On the other side, functions like deduplication and backup of course can be added to EVO:RAIL with vSphere Data Protection (VDP) or third-party tools from vendors like Veeam, Symantec, EMC or NetApp. However, the unique selling point for SimpliVity is the integration of these functions within a hyper-converged product.
EVO:RAIL and vSAN taking on storage vendors
Even as the possible use cases are a little constrained and functions like deduplication and backup are missing, EVO:RAIL seems like a capable alternative to converged or hyper-converged products from vendors like SimpliVity, Nutanix, NetApp or Cisco. Through the various OEM partners (Dell, HP, Fujitsu, Hitachi Data Systems, EMC, NetApp, Inspur, NetOne and Supermicro) it´s highly likely that VMware is capable of establishing a broad ecosystem of different EVO:RAIL appliances for different use cases and even more demanding workloads.
Another remarkable aspect is the impact of EVO:RAIL on the storage market. Tony Parkinson, who works at Dell, says EVO:RAIL is a serious threat for storage vendors. This already has been the case with VMware vSAN and it becomes more of a threat with the integration of vSAN into EVO:RAIL. At VMworld Europe, many storage vendors discretely were deeply concerned about EVO:RAIL combined with VMware’s market power.