A probe into the Nutanix hyper-fusion system

Source: Internet
Author: User
Tags isilon

referring to the technological trends of storage systems, it is generally considered that hyper-convergence and Server SAN is the trend of the future mainstream, will replace the traditional storage technology architecture. So what is the technical uniqueness of hyper-fusion and Serversan ? Why can it be a mainstream trend in the future of data center storage architectures? What exactly does this architecture solve for data center problems? Where is the technical difficulty that this architecture needs to address? Here, the storage of old Wu and everyone together to analyze the Nutanix system, from the Nutanix system to see this hyper-converged system architecture unique, and can experience the hyper-fusion and Server SAN the essence of the.

Nutanix is by Google A storage company created by engineers, and many of the people in this team are GFS Engineer, the company draws on the internet's distributed experience, builds a storage system on the basis of a general-purpose server, and integrates storage and computing into a hyper-converged system, specifically addressing storage issues in virtualized computing environments. The company has gone through 5 rounds of financing, with a total of more than $317million in financing. While facing Vmware 's VSAN Impact, there is a very broad market outlook.

Nutanix There are too many concepts in this system, such as software-defined storage, hyper-converged, Server SAN and so on. Behind so many miscellaneous concepts, we first need to clarify What the Nutanix needs to solve the storage domain problem? Storage Lao Wu thinks that Nutanix needs to solve the storage problem in the virtualized computing environment. Let's take a look at the system architecture in a traditional virtualized computing environment, as shown in:

650) this.width=650; "title=" 1.jpg "alt=" wkiom1xpmf2tierdaadnokduk48217.jpg "src=" http://s3.51cto.com/wyfs02/M00/ 72/a3/wkiom1xpmf2tierdaadnokduk48217.jpg "/>

the traditional virtual computing system consists of a blade server or a common server node to form a computing cluster. By deploying virtual machines on these actual servers, the deployment of virtual machines can simplify the management of compute resources. Throughout the system, all virtual machines share centralized storage through the back-end storage network. These centralized storage can take EMC 's VMAX,VNX , or Isilon storage systems. To improve the high availability of storage services, storage networks typically require redundancy, requiring at least two storage switches (FC switches) to connect compute and storage resources. The biggest problems with this traditional storage architecture are the following:

1, the centralization of storage resources, it is difficult to expand with the expansion of computing resources. In particular , the storage system, which uses the Scale up of VNX, cannot adapt to the scalability requirements of the data center.

2, the use of storage and computing separation of this architecture, the need for independent network, independent storage and independent computing. In order to improve the usability of the system, it is necessary to achieve the design goal through resource redundancy. Therefore, the network, storage controller (dual control) will be redundant, increasing the investment in the data center.

3, the deployment of many different systems in the data center undoubtedly increases the management complexity of the data center. In particular, multiple types of storage devices at the backend increase the cost of managing the data center.

We know that the trend of data center is virtualization, software-defined data center is the direction of development. Data centers based on a variety of complex devices will eventually be normalized to the architecture of the Universal Server plus Interconnect network. Deploy critical software on these general-purpose servers and implement compute and storage resources in a virtualized way. Then deploy the application on the basis of these virtualized resources to accomplish the specific functions.

650) this.width=650; "title=" 2.jpg "alt=" wkiom1xpmjyjngjnaafgpxrqh7g613.jpg "src=" http://s3.51cto.com/wyfs02/M01/ 72/a3/wkiom1xpmjyjngjnaafgpxrqh7g613.jpg "/>

driven by such a development trend, Nutanix this hyper-converged system has been launched to complete the task of distributed storage virtualization for virtualized data centers. In line with the data center to the cloud era of the development of a technology trend.

650) this.width=650; "title=" 3.jpg "alt=" wkiol1xpnhqysiecaaeeg4_na74567.jpg "src=" http://s3.51cto.com/wyfs02/M02/ 72/9f/wkiol1xpnhqysiecaaeeg4_na74567.jpg "/>

Nutanix the system architecture for hyper-converged systems is as follows:

650) this.width=650; "title=" 4.jpg "alt=" wkiol1xpni_jzmocaaeuq5x7yh4182.jpg "src=" http://s3.51cto.com/wyfs02/M00/ 72/9f/wkiol1xpni_jzmocaaeuq5x7yh4182.jpg "/>

This is a typical Scale OutDistributed storage architecture. NutanixThe system is composed of a series of server nodes, connecting these nodes through the back-end network, forming a set of distributed computing and storage System. In eachNutanixThere are multiple virtual machines in the node where at least oneController VM, the virtual machine is responsible for the management of local storage resources and implements a distributed storage system in conjunction with other nodes. Objectively speaking,Controller VMThe distributed storage andEMCof theIsilonis similar,Controller VMRun inNutanixA core Distributed file system that virtualize local and Remote Storage resources andSSDtheHDDoptimize for performance and achieve tiered storage. Controller VMthe external can provide traditionalISCSI,NFSand other interfaces. Because the applicationVMSand theController VMare located in the same physical device, so twoVMS(ApplicationVMSwith theController VMcan also be used to achieve more efficient storage access through a special interface.

simply from the perspective of this architecture,Nutanixsimply running the storage operating system into theVMSInstead of the traditional controller (physical server). EMCof theIsilonwith the actual physical server to implement the storage function, from a technical point of view, there is really no fundamental change. Migrating storage capabilities to virtual machine platforms actually makes storage development easier. The virtual machine platform normalized the different physical server platforms, allowing the storage vendor to transparently process the physical server platform and simplify the design. Backup leader of the yearDatadomainThis is the way of thinking. DatadomainMainstream products are implemented on a physical server platform, installing the core on a custom physical server platformDatadomainthe operating system implements data deduplication for backup data. In virtualized applications,Datadomainalso provides a virtualized solution to install on a virtual machine platformDatadomainThe operating system implements a virtualized deduplication system. This backup VM can be on the same physical server platform as the application virtual machine, or the backup virtual machine can be installed inVMAXin such a high-end array, the realization of the integrated machine solution. So, from this point of view,NutanixIn fact there is nothing extraordinary in the architecture,EMCof theIsilonthe platform can be easily packaged into a set ofNutanixsimilar hyper-converged systems.

implementing storage capabilities on virtual machine platforms is essentially handing over a large portion of server platform management to virtual machine vendors, such as KVM or Vmware , thereby alleviating the workload of the storage vendor. Storage vendors need to study a variety of server platforms, and now only need to study a few vmware or KVM Limited platform. However, everything is two-sided, because the virtual machine software itself is still a lot of defects, itself is still in the process of development, so, after migrating to the virtual machine platform, not necessarily to the storage vendors to reduce the workload. This we can see from The Nutanix team composition, the virtualization of its engineers accounted for a large proportion. Although the server platform engineer has been reduced, it has increased the resources of engineers in the direction of virtualization.

So, storage old Wu thinks that Nutanix There is not much innovation in the architecture of the product, there is such a figure in traditional storage. Nutanix An important success in meeting the trend of data center development, making great innovations in hyper-converged products, and proposing to completely abandon San/nas, accelerate the data center cloud, The process of virtualization. This is a huge innovation in solutions and products.

NutanixThe core of the product is a set of software that is callednutanixoperating System(NOS), inNOSThe core of this is the Distributed file system. The file system implements the distributed storage function, which willSSDand disk resources to be globally virtualized and then organized into volumes or files for other application virtual machines. From the perspective of applying a virtual machine, it can beISCSI Initiatorthe way to mount the distributedController VMThe volume that is provided. Data is redundant between physical nodes and can be used in multiple copies orErasure Coderedundancy data, thus guaranteeing the reliability of the data. Storage Virtualization Technology-basedNutanixThe schema looks like this:

650) this.width=650; "title=" 5.jpg "alt=" wkiom1xpmowbjyaaaagwpen5lf0407.jpg "src=" http://s3.51cto.com/wyfs02/M00/ 72/a3/wkiom1xpmowbjyaaaagwpen5lf0407.jpg "/>

can sayNutanixis an outright software-defined storage system. With the help of virtual machine technology, storage can be deployed to any common server platform, soNutanixis also aServer SANSystem. From a technical point of view,Nutanixis also a tiered storage system thatNutanixwill be equipped with the inside of the nodeSSD, theSSDused as a hot data cache, all hotspot data can be placed inSSDthe. In theNutanixInterconnection between nodes through an interconnected network, the interconnection network can be10GbEthernet, or it can beIBNetwork. Under such an architecture, the performance bottleneck of the entire system is not the storage disk, not theCPUcompute resources, and more in the back-end interconnect. In the process of data access, both read and write operations require data transfer through the back-end network, which can affect the overall throughput of the request andIOdelay. In particular, in the presence of distributed caches in the system, this architecture can cause problems with the scalability of the system. We can take a look at this .Vmwareof theVSAN, why it has been difficult to extend the number of nodes linearly, the main reason is that there is data correlation between nodes, and there is a problem of distributed lock. From a technical point of view, in the field of storage has been studying this problem, design a large scale Distributed file system/DistributedSANsystem is the technical difficulty lies. At this pointNutanixadopted aLock-freeconcurrent processing model, with the design of lock-free concurrency greatly improve performance, can achieve linear Scale Outthe purpose. As shown,Nutanixin theVDIlinear and scalable performance under application:

650) this.width=650; "title=" 6.jpg "alt=" wkiol1xpnmxdutrdaad2d4nuafo485.jpg "src=" http://s3.51cto.com/wyfs02/M02/ 72/9f/wkiol1xpnmxdutrdaad2d4nuafo485.jpg "/>

NutanixAnother important feature is the provision of highly reliable data services and highly available application services. When there is a problem with the application server, you can take advantage of the virtual machine-providedVMSThe Migration Service, which migrates applications from the failed node to an available server node, enables the application to be switched over a short period of time to ensure high availability of the service. When aNutanixwhen a local data disk in the node fails, theController VMwill take over the functions of the data service and restore the data through distributed data redundancy to ensure the reliability of the data. BecauseNutanixThe data management method of the storage pool is adopted, so that the whole systemController VMcan be combined to recover data for the fault disk, so that the data can be recovered in a very short period of time, ensuring the reliability of the data. This idea and Huawei'sRAID2.0are essentially the same. JustNutanixAdoptErasure CodeData redundancy level or multiple replicas to achieve data protection, is essentially a distributedRAID. In terms of high availability of services,NutanixTake advantage of the virtual machine'sVMSmigration functionality, which is well worth learning.

Another direction of storage is that flash technology, especially the drive and development of all-flash systems, will lead the data center to all-flash. In nutanix nvme SSD rdma roce 100GB nutanix This platform can not fully play the nvme SSD performance, This is also a problem with this system. Flash performance is not good, which means nutanix The future of such platforms may be more appropriate for applications that are not particularly sensitive to performance, and that are very important for scalability and manageability.

See Nutanix after such a computation and storage fusion architecture, storage old Wu was thinking, if in Nutanix All-flash technology is introduced on the platform, a Nutanix How many nodes are needed CPU , memory resources? In an all-flash system,theCPU itself is a valuable resource, an important performance bottleneck, and a large amount of memory resources need to be used in flash storage systems. Under such resource demand, how to partition a part of the resource running application service in this all-flash system? How do I do this, how can I ensure the performance of my application and the consistency of my IO performance? This kind of control will become extremely complex. Therefore, in the massive all-flash system,Nutanix This hyper-converged architecture is difficult to achieve low latency, consistent performance.

Unlike compute and storage-integrated hyper-converged architectures, another high-performance, independent software-defined distributed storage System with virtualized compute nodes can also build software-defined virtualized data centers, which can be equally scalable and manageable with a separate architecture for storage and computing. and can achieve good performance, application network and data network can be configured according to the specific needs, but also can bring a good overall investment costs. The North American Solid Fire provides the data center with an all-flash system that provides high-performance storage services for the data center. So, will the future of virtualized data centers Nutanix architectures unified the lake? Personally, I think that's not very realistic. However, distributed storage must be the development direction of the virtualized data center, and the traditional dual-control San/nas will not adapt to the development direction of data center cloud.

Overall, Nutanix is a very advanced hyper-converged system that leverages the virtual machine platform to implement distributed storage on a common server platform in a software-defined way, completely discarding the traditional San/nas System. Reduces the overall cost of data center management and equipment investment. From the point of view of technology, The key point of Nutanix system is Distributed File system, which realizes a set of distributed file system which is easy to be extended. With the development of the cloud data center, I believe that the Nutanix system will be ambitious.

This article is from the "Save the Way" blog, make sure to keep this source http://alanwu.blog.51cto.com/3652632/1691407

A probe into the Nutanix hyper-fusion system

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.