It is incredible that it already more than a year since my last post here at VDIcloud, and it’s even worse that I promised more content but never delivered. So let’s give it another try, and where better to start than with the Tech Field Days I’ve been to in the past. Let’s start with the last I’ve been to (also almost a year ago) at the NetApp Insight event in Las Vegas in 2018. As the next event will be this month, I thought it would be a good idea to see what they showcased last year, and then after the event this year compare and see what NetApp brings to the table.
How to give HCI power to the customer
We live in a world where data is worth more than oil, and companies are constantly looking for better and faster ways to utilize their data to the fullest. A lot of times we here about cloud, and instantly we think about Microsoft Azure, Amazon Web Services and the Google Cloud Platform, but a lot of companies are still working on their own on-premises infrastructures albeit that more and more companies are trying to replicate the public cloud providers by utilizing Converged and Hyper Converged solutions. Bridging the gap between On-premises solutions and Public Cloud Providers is something that a lot companies are struggling with and for this they are looking at the HCI providers for help. NetApp is relatively late with their HCI solution, but that doesn’t mean that they’re providing the right solution for their customers.
NetApp’s view on HCI
During this presentation Gabriel goes deeper into what the focus is for NetApp HCI and provides answers on some of the important questions NetApp’s view on HCI. The presentation was focussed on what defines HCI, what types of HCI exist, and what are the benefits for customer. As always with the Field Day events there are some awesome delegates that have great questions an remarks on what the presenter is telling, and this presentation has some great remarks on HCI, CI and what the difference really is. I really loved the conversation on what the difference is between CI and HCI and what that means for the customer, but I do agree with Gabe that the biggest difference between CI and HCI is that CI comes by the rack and the stack is controlled by the vendor, where as with HCI you can start small and grow big. Both have their advantages, but if convergence started with the hyperscalers, HCI is much more in line than CI. NetApp offers a solutions that exists of Compute nodes and Storage Nodes that can be scaled independently of each other. This provides customers the ability to really disaggregate compute and storage, while still having the power of a single scalable and easy to maintain platform.
I’ve always looked at the HCI solution as a great step towards building on-premises cloud solutions (private cloud if that is a better name for you to use). The problem I’ve always seen with these solution is that the focus often seemed to be on the bringing the vendors solution to the customer and in doing so it is not always bringing the best solution to them. Hyper Converged Infrastructure for me is much more about bridging the private and public Cloud and for me the solutions that NetApp provide to its customers seems to meet these standards. So calling in Hybrid Cloud Infrastructure instead of Hyper Converged is a great one. Sure there is work that needs to be done by NetApp, but who knows what they will announce in a couple of weeks. Just watch the video and read the blogs below to have the best insights on the NetApp HCI solution. And don’t forget to visit the Tech Field Day Xtra site for this event here: https://techfieldday.com/event/netappinsight18/
I just wanted to shout out to my fellow TFDx delegates at this event that wrote some great insights to this preasentation:
Tuesday 18 april 2017, Patrick van Helden, Director of Solution Architecture at Elastifile was at Metis IT to tell about Elastifile. We had the chance to try a real-life deployment of the Elastifile software.
Elastifile is a relative new name in the storage area. Since this month, the company is out of stealth and has presented its Elastifile Cloud File System. The company is founded in 2013 in Israel by three founders with a strong background in the virtualization and storage industry. In three funding rounds the product raised $58 Million. In the last round $15M came directly from Cisco. Other investors in Elastifile are leading flash Storage vendors and Enterprise Cloud Vendors.
What is Elastifile?
The goal of the founders is to have a storage platform that is able to run any application, on any environment, at any location. Whereby any location means really any location: Cloud or on premise. The product is developed to run with the same characteristics in these environments. Therefor Elastifile wrote from scratch a POSIX compliant filesystem that supports file, block and object oriented workloads and is optimized for flash devices. You can store your documents, user shares, VMware VMDK files, but also use it for big data applications, all stored on the same Elastifile Cloud File System.
But what is the difference with a NetApp storage for example? A NetApp system can also provide you the same capabilities and is already doing this for years. The first thing in which Elastifile’s approach is different than NetApp, is the way the product is written. It’s written for high performance and low latency. Elastifile only supports flash devices and the software knows how to handle the different types of flash devices to get the best performance and extend the lifetime of flash devices. Furthermore, ElastiFile is linearly scalable and can be combined with compute (Hyperconverged Solutions).
Another difference is that the Elastifile Cloud File System can run inside a (public) cloud environment and connect this to your own on premise environment. The problem with (public) cloud environment is that it gives you not the same predictable performance as in your on-premise environment. The Elastifile Cloud File System have a dynamic data path to handle noisy and fluctuating environments like the cloud. Due to this dynamic path Elastifile can run with high-performance and most important with low latency in cloud-like environments.
Elastifile’s Cloud File System can be deployed in three different deployment models:
Dedicated Storage mode
The first deployment model is HCI, where the Elastifile software runs on top of a hypervisor. Now, Elastifile supports only VMware, additional hypervisors will be added in future releases. You can compare this deployment with many other HCI vendors, but when connecting and combining the HCI deployment model with one of the other deployment options it gives you more flexibility and capabilities. Most other HCI vendors only support a small set of certified hardware configurations, wherein Elastifile supports a broad range of hardware configurations.
The second and in my opinion the most interesting deployment model is the dedicated storage mode deployment. In this model, the Elastifile software is directly installed on servers with flash devices. Together they create the Elastifile distributed storage. With this deployment model, it is possible to connect hypervisors directly to these storage nodes using NFS (and in the future SMB3), but also connect bare-metal servers with Linux, Oracle or even container based workloads to this same storage pool.
As we already discussed earlier the latest deployment is the In-Cloud deployment. Elastifile can run In-Cloud in one of the big public cloud providers but is not limited to public clouds. Elastifile can also run in other clouds as long it delivers flash based storage as infrastructure. The Elastifile can use the storage to build its distributed low-latency cloud file system.
When combining these three models you get a Cloud ready file system with high performance, low latency and a lot of flexibility and possible use-cases.
HCI file services
A great use-case for the Elastifile Cloud File System is that you can decouple the operating system and application from the actual data of the application in a HCI deployment. You can use the Elastifile Cloud File System to mount a VM directly to the storage and bypass the hypervisor. And because the Elastifile Cloud File System is a POSIX filesystem it can store millions of files with deep file structures.
Linear scalable in cloud-like environments
A second use-case for the Elastifile Cloud File system is that the performance with any deployment of Elastifile delivers a predictable low-latency performance. When expanding the Elastifile nodes each node will add the same performance as any other node. When adding additional storage, you’re also adding additional storage controllers to the cluster. This result in a linear scalable solution even in cloud-like environments.
The last use-case of the Elastifile is that it could automatically move files on the filesystem to another tier of flash storage. This could be a cheaper type of flash or a less performing type of flash storage, for example consumer grade SSD’s. Movement will be based on policies. The Elastifile software can further offload cold data to a cheaper type of storage, like a S3 storage. This can be a cloud based S3 storage, but can also be an on premise S3 storage.
How the future will look like is always difficult to say, but from all what I already tried is this a very promising first version of the Elastifile Cross-Cloud Data Fabric. In the session with Patrick, I deployed the software myself and Patrick showed us the performance on these deployed nodes without any problems. The idea’s around the product are great and on the roadmap, you find the most important capabilities which are needed to make it a real mature storage product.
This is a cross post from my Metis IT blogpost, which you can find here.
After teasing the market with a photo containing three servers, the word Hyper and some blank puzzle pieces, Cisco announced their own Hyper-converged Solution: Cisco HyperFlex. This solution is an extension of Cisco’s Unified Computing System (UCS). Until now the UCS platform portfolio did not contain a native Cisco storage solution. Finally Cisco entered the highly competitive Hyper-converged Infrastructure (HCI) market with HyperFlex.
The Cisco HyperFlex solution combines compute, storage and the network in one appliance. Cisco says the solution is unique in three ways: flexible scaling, continuous data optimization and an integrated network. All other HCI vendors do Hyper-converged with compute, storage and networking, but none of these have a complete integrated network solution. As expected of a former networking only company, Cisco also integrated the network.
The platform is built on existing UCS components and a new storage component. The servers used in the solution are based on the existing Cisco UCS product line. Networking is based on the Cisco UCS Fabric interconnects. The new storage component in Cisco’s platform is called the Cisco HyperFlex HX Data Platform, which is based on Springpath technology.
Springpath HALO and Cisco HyperFlex HX Data Platform
Springpath was founded in 2012 and Cisco co-invested the start-up. Springpath has developed its own data platform using HALO (Hardware Agnostic Log-structured Object) architecture. The HALO architecture offers a flexible platform with data distribution, caching, persistence and optimization. Cisco has re-branded this to the Cisco HyperFlex HX Data Platform.
All data on the Cisco HX Data platform is distributed over the cluster. Data optimization takes place by using inline de-duplication and compression. Cisco indicates most customers should reach 20-30% capacity reduction with de-duplication and another 30-50% with compression without any performance impact.
VMware and Cisco HyperFlex
First the HyperFlex solution will only be available with the VMware hypervisor using NFS as storage protocol. A Data Platform Controller for communication with the physical hardware will be used for the HyperFlex platform. This Data Platform Controller requires a dedicated number of processor cores and dedicated amount of memory. The controller integrates the HX Data Platform with the use of two preinstalled VMware ESXi vSphere Installation Bundles (VIBs): IO Visor and VAAI. IO Visor provides a NFS mount point and VAAI offloads file system operations.
The HyperFlex storage is managed with a vCenter plug-in. There are currently no details available about the layout and functionality of this plug-in. We expect the plugin will be the same as Springpath with Cisco branding.
The physical server and network is managed like any other Cisco UCS server. Each server will be connected to the Fabric Interconnect and managed from the UCS manager interface.
Cisco HyperFlex range
The HyperFlex platform is available in three different models, an 1U and 2U rack based server and a combination of rack servers with blade servers. The first model is for a small footprint, the 2U model is for maximal capacity and the last option is for maximal capacity and high compute.
All configurations must be ordered with a minimum of four servers. As far as we know at this stage the maximum number of servers in a HyperFlex cluster is eight. Each server will be delivered with VMware pre-installed.
The hardware configuration of the HyperFlex nodes is not fixed. You can choose your type of processor, amount of memory and the amount of disks. On the Cisco Build & Price website all available configuration options can be found. You can always scale your cluster by adding storage and/or compute nodes.
Cisco has an interesting licensing model for the HyperFlex HX Data Platform. The HX Data Platform will be licensed on a per year basis. In the configuration tool by default a server is configured with a license for one year. This licensing model deviates from other HCI vendors who base their license model on raw or used TB’s, or use a perpetual license.
Cisco is a new and interesting player in the rapidly growing Hyper-converged market. The technology used provides some nice features, capabilities and an interesting licensing model. Time will tell if the product will be successful and what the roadmap will bring for the future. But at first sight it looks like a good alternative for the leading Hyper-converged solutions.