In-depth analysis of the converged Architecture

Source: Internet
Author: User

Author: Fan Jun Frank Fan) Sina Weibo: @ frankfan7: frankfan7

In the previous article, we introduced the basic concepts of the converged architecture. This article will show you how to explore the xuanjicang step by step.

Not to mention the traditional server giants, a new leader, Nutanix, caught my attention. Market performance is not the focus of this Article. Let's talk about technology directly. My analysis is based entirely on the vendor's information and I have no chance to test it myself.


650) this. width = 650; "src =" http://img1.51cto.com/attachment/201305/154818818.png "title =" nutanix (mobileapps.png "style =" float: none; "/>


Is a 2U Block with four servers integrated.


I think Nuntanix has the following highlights:

ILow cost and high performance

With the low cost of local storage, it can achieve the high performance of SAN. That is to say, it costs less, but the performance is not bad at all. Isn't that a blow? Let's take a closer look.

Shared storage has always been recommended in virtualization design. Because shared storage is not available, vMotion, HA, DRS, and other features cannot be used. However, in recent years, local storage has gradually become a stage in virtualization design,FirstBecause the cost is much lower than shared storage such as SAN,SecondRelatively simple management and fast implementation,Third,The software layer can use local storage to implement a distributed file system for sharing, redundancy, scalability, and high performance.


650) this. width = 650; "src =" http://www.bkjia.com/uploads/allimg/131228/0141245601-1.png "title =" n5 (mobileapps.png "style =" float: none; "/>

VMotion

650) this. width = 650; "src =" http://www.bkjia.com/uploads/allimg/131228/0141246213-2.png "title =" n3 (Mobile).png "style =" float: none; "/>


Taking the VM on the first host as an example, each write operation is written to the local storage and written to the local storage of other hosts managed by the same primary Controller to achieve redundancy.

Assume that vMotion ranges from the first host to the third host, the Controller checks whether the third host has the storage information of the VM and finds that the red and orange parts are missing, then, the missing parts are copied to the local storage of the third host. Then start vMotion.


High AvailabilityHA)

650) this. width = 650; "src =" http://www.bkjia.com/uploads/allimg/131228/014124E39-3.png "title =" n2 (Mobile).png "style =" float: none; "/>

At least three servers are required for high availability. Each Server includes an operation, storage, and virtual layer. The command line ControllerVM of the entire solution is completely implemented by software.

The Controller on each host communicates with other controllers, one of which is the master and the other is the slave. When the primary Controller goes down, a Slave Controller automatically becomes the primary Controller.

Take the VM on the third host as an example. Each write operation is written to the local storage and written to the local storage of other hosts managed by the same primary Controller to achieve redundancy. If the third host goes down at this time, the Virtual Layer HA will restart the vm on the second host because a backup of the VM is available in the local storage of the second host.


IIScalability and predictable performance

The Design of SAN must take into account factors such as IOPS and Throughput that can be supported by Optical Fiber switches, StorageProcessor, and other nodes. If it is not planned to take weeks, SAN capacity expansion may cause performance degradation due to a bottleneck at a specific location.

The extension of Nutanix looks much simpler. Local Storage and computing are integrated, and you only need to increase the number of hosts for expansion.

650) this. width = 650; "src =" http://www.bkjia.com/uploads/allimg/131228/0141246456-4.png "title =" n4 (Mobile).png "style =" float: none; "/>

According to data provided by Nutanix, the performance after expansion increases linearly, which is particularly suitable for the practice of desktop virtualization. This means that you can use a small investment to test it in the initial stage. When the number of virtual machines increases significantly, the number of hosts increases, and the performance is the same as before. There is no performance bottleneck.


Three storage layers

650) this. width = 650; "src =" http://www.bkjia.com/uploads/allimg/131228/014124F52-5.png "title =" n6 (mobileapps.png "style =" float: none; "/>

As you can see, local storage is controlled by the software layer and shared to all hosts in iSCSI/NFS mode. Each time a virtual machine reads and writes shared storage, the Controller is used for scheduling. Common data is processed by the fastest FusionIO. The data that is not commonly used is still in the local hard disk.


Question:

Some questions require practice and further verification:

FirstBecause each Write IO initiated by a VM must be stored locally on multiple hosts. This is a Transaction. after all transactions are successful, the Write operation is valid. If many nodes are controlled by a Controller, they do not know the impact on performance.

SecondController stability also needs to be verified. When the primary Controller is down, can other slave controllers seamlessly take over the cluster? What is the switching time?

ThirdThe high availability implemented by Nutanix in a data center is no problem. However, when a large enterprise needs disaster recovery between two data centers, data synchronization between data centers is often involved. I don't know how to synchronize the local data stored in Nutanix in different regions.

My theoretical analysis is based entirely on the information provided by the manufacturer, hoping to see more data in the real environment in the future. Please pay attention to the following further analysis on the Controller and the file system of Nutanix.

This article from "sit up and watch the cloud" blog, please be sure to keep this source http://frankfan.blog.51cto.com/6402282/1198017

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.