Reprinted from "Network World"
Original link:
Http://www.cnw.com.cn/issues/Article.asp?filename=n43630.asp
Linux has been growing, this is an indisputable fact, however, any growth is accompanied.
Linux has been cultivated by some open source enthusiasts, some people who have been sought after by the underlying technology and use the open source Linux, which build some unique applications, people hope that such an operating system can break some proprietary Or the system that forms a monopoly situation has always been overbearing, they think that Linux should enter enterprise-class applications, especially enterprise-class applications, which can get real use value, and growth can be transformed into wealth. Labor. This is the reason why more and more Linux commercial versions of system vendors are also the cause of more and more related fields to accelerate their transformation into labor. From the late 1990s, the Linux system manufacturer, IT giant crocodile, how is the enthusiasm of Linux, but what is the user's view that it is transformed into a labor force? What kind of troubles have been experienced in the process of Linux growing to real labor? I hope that everyone can find some answers from this topic.
I recently made a small-scale user survey on our CNW website (www.cnw.com.cn), and we believe that the results of the survey have considerable reference value for users, and can give some answers to the above problems.
Present article
Experience the forefront of Linux application
Research Problem 1: Is the Lee I'm of the Linux operating system?
Result data:
Data Enlightenment: The user's familiarity of Linux is a very important issue, familiarity, in a sense, determines their accepted level. The result data shows that most users are only "contact" for Linux, accounting for 37.5% of the effective survey results, from a more in-depth telephone access, the reporter learned that this "contact" includes some of the test experience, including With our own experience, including some propaganda, training, etc., basically staying a very preliminary understanding, but some features of Linux are unclear. Users who are very familiar with Linux generally R & D or maintenance personnel, and their daily work is developing or maintaining Linux applications such as Linux-based web servers.
Research Problem 2: Viewing Linux as an enterprise operating system?
Result data:
Data Enlightenment: In fact, in some SMEs, Linux is a very common phenomenon as an operating system for deploying a web server application. In the previous year, we "Network World" evaluation laboratory organization-based web server testing pre-survey I learned that the SMEs with nearly half are selected in the deployment web application. This survey is to understand the user's views for using Linux as an enterprise-class business application platform, so we have four contents. For specific alternative answers.
From the findings of the survey, no one thinks that Linux is still immature. Most still believe that the Linux system itself has complete, mature, but lacks application software, that is, lack of application is the biggest defect in Linux. It is basically the basically developed by Linux, and some integrated engineers, they have helped customers have deployed Linux-based applications. In an interview with some hardware system vendors and middleware vendors that support Linux platforms, the reporter learned that the cost of the cost is mainly derived from the server hardware products of the IA architecture and the cost of the Linux operating system itself. Users start to recognize and Accepted, however, this application is mostly limited to Linux applications the most mature web server applications. There are also very few users to start using Linux-based database and deploy application server software, but relative to other system platforms, especially Unix, Windows platforms that deploy key business applications, using Linux still a few. Part of this is part of Linux's immature, the system itself is not complete, more importantly, lack of application. Not long before UNISYS has just announced that its ES7000 server can run Linux, Unisys and Novell, Red Hat work together to provide services and support based on Linux solutions, and they also publish the first dynamic partition feature in the Linux system. Chen Zhongli, general manager of UNISYS Greater China System Science and Technology, is mentioned in an interview with a reporter that although this cooperation is increasing, there is more and more hardware platforms that support Linux, but relative to Windows platforms, Linux applications Too few, this takes time and more cooperation between system vendors, ISV, etc., he believes that Linux has taken at least 2 to 3 years. In addition, 62.5% of 62.5% have chosen the lack of application software in the "reason for deploying Linux", which also shows that the application is the key to Linux enterprise applications.
Research Problem 3: What kind of application is deployed by the Linux system?
Result data:
Data Enlightenment: In this question, most of the users use one of the choices to deploy the web server, this ratio reaches 75%. It is considered that Linux can serve as 22.5% of the platform for deploying database servers and large application software / critical business applications. It is considered that the proportion of users who can be used as a database server and deploy application server software platform is also large. There are also some surveys on the Internet show that Linux-based open source database is increasingly concerned, and the challenge of the market comes from three major competitors: mysql, maxdb and postgres sql. In fact, deploying large database servers is also one of the enterprise-class applications. From these data, users hope that the scope of the task that Linux can bear has expanded, and slowly close to the key field, this is in the system manufacturers. The reason for advancement. Based on the LINUX system, it is also vigorously promoted the application and development of Linux itself, including the development of the Nuclear grade, and the new database products are developed for some commercial Linux systems. Although It has a mobile phone, which is the movement of the operating system, Jiangshan, but as one of the most important database manufacturers, is still quite large. Unisys believes that the IA architecture is combined with the flexibility and openness of the Linux operating system, which provides users with new products with high performance and price, in the field of high-end applications, can replace prices expensive RISC-based proprietary systems. When UNISYS, Chen Zhongli, told reporters that now, it has now arrived in the UNIX / RISC market, and the application field of traditional RISC systems should be the breakthrough point of Linux to high-end key business applications. It is worth noting that users from insurance, electricity, etc. have no choice to deploy Linux to run a key business. In a more in-depth telephone interview, they are still conservative for this problem. On the one hand, the familiarity of Linux is not enough, and the other side is more doubtful to Linux security and functionality.
Research Problem 4: What are the efforts of Linux as an enterprise core application system?
Result data:
Data Enlightenment: From data, the support of application software is the most prominent issue. As can be seen from the "Application Software Lack" diagram, it is considered that the lack of professional application software reaches 72.5%, and it is considered that the proportion of users who lack large management software and application server software exceeds 50%. This is to say, build Above Linux, professional software for special applications is urgently needed, and independent Linux manufacturers should increase the cooperation with ISV, integrators, develop targeted industry solutions and professional applications; management software and security classes Software vendors should also see such business opportunities, accelerate support for Linux and cooperation with Linux system vendors.
There is also a prominent problem is the support of technology, service. In a telephone interview, many users feel that the service of Linux system manufacturers is not in place. Although it is an open source system, once as a commercial system is purchased by the user, it should provide adequate technical support and services, and users generally want to provide on-site service. To solve the problems they have encountered in their applications, some senior technical engineers need technical consultation or training.
Some users have referred the shortcomings of security. They believe that Linux does not have too many users like Windows, and finding what security issues can be remedied in time, that is, Linux still lacks some security mechanisms and timely security. The remedy is impossible to prevent some types of attacks, so Linux urgently needs more security software support.
The cost of Linux has always been the strongest weapon to promote him to enterprise core applications. Whether it is system manufacturers, database manufacturers or middleware manufacturers, Linux is replacing Windows and proprietary Unix The free alternative products of the old-name operating system, especially if Linux combines with the IA architecture hardware system, and its cost advantage is more prominent. But the fact is that everything has a price, including open source software, and the middleware built on Linux is also good, the database will not become cheap because it is Linux. Here we only look at Linux itself, calculate buy, migrate, operate, and support costs, is it really cheap? In other words, are Linux's TCO (total cost) lower than Unix or Windows TCO? Middle articles
True cost of Linux
The simple answer is as follows: Enterprise uses Linux in its IT infrastructure, the more comprehensive, it may get more economic benefits from the previous investment of this operating system. These investments include Linux training and tools and fees from UNIX or Windows environmental migration, economic benefits will have better management tools, more third-party manufacturers support and more technical expertise. keep growing.
The benefits from UNIX or Windows to Linux are driven by 4 primary costs: purchase, migration, management, and support costs. Some key issues in various fees are briefly discussed below.
Hardware and software costs
The cost savings that use large-scale production IA architecture machines replace expensive RISC processor-based Unix systems are one of the most important factors that use Linux. The calculation hardware cost is easy. Forrester Research Chief Analyst Ted Schadler said: "Due to the economic performance of IA, it is easy to make a very easy decision to make a decision on Linux, the hardware of the RISC architecture and the hardware of the IA architecture The difference in system prices is between 5,000 to 25,000 US dollars.
In terms of software, the cost difference is not obvious. The study of IDC, Meta Group, Robert Frances Group found that Linux licenses were lower than the WINDOWS license fees. But some analysts believe that this is not a comprehensive one-on-one comparison. Forrester's Schadler pointed out: "When developing applications, choose Windows or Linux is not a problem, this is the Java selected on Linux or the decision of Windows. Microsoft bundles a lot of things in Windows, SQL Server, and .Net framework, if planned An ordinary application and deploying it with all the prices, Windows will win unfillly, because it will get so much bundled things. "
In terms of hardware and software, the cost advantage of Linux is often overlooked is its flexibility provided on future migration and upgrade roads, because when using Linux, you can master your own upgrade cycle.
Migration fee
When considering migrating from UNIX or Windows to Linux, companies should carefully analyze one-time migration fees. One of the biggest costs is that training system administrators master Linux as soon as possible.
In fact, UNIX skills are very close to Linux skills, which reduces the cost of migrating from UNIX Linux. For a pure Windows user, there is no Linux skills, so obstacles are very huge. Those who have UNIX skills but cannot be adapted immediately, can easily download Linux, trial at home.
Since the re-training cost is minimal and the hardware procurement cost is reduced, UNIX to Linux migration is generally economical; however, due to re-training costs and less hardware purchase costs, Windows to Linux migration is more like a piece Things to decide by luck.
Other migration costs include code that may have to be overwritten in order to reproduce Linux intended platforms, must be overwritten, and work with backend systems and software that must be purchased; depreciation is not expected. The hidden migration fee; some middleware, such as the application server, can be easily ported to Linux, and if the original application has been running on such a middleware, a one-time migration fee will be reduced. Management fees
So far, the maximum cost of most Linux TCO research is the fees required by future operational system management. In the 2002 IDC, a comparative study of Windows and Linux is 62% of these two environment for 5 years. It has also caused the largest cost difference between the two systems, and Windows leads the first step in a slightly low cost. .
Although most analysts believe that Windows managers are lower than the Linux system administrator, but the actual debate focus focuses on integrated cost gains and the availability and quality of management software tools for Linux platforms.
Oracle's view is that users who simply use Linux with other operating systems may not save money, and put forward the argument: If you fully migrate to Linux and buy a strong management tool, then on Linux Save more costs, because Linux developers can use the Linux operating system to manage more machines, larger workloads, but only require less managers.
Some large Unix systems believe that on Linux, tools and processs have not developed to their levels achieved on UNIX, so the cost of migrating to Linux saves is zero.
A medical center CEDARS-SINAI chief technology officer Duncan found a key tool for some cost-saving, although he was satisfied with its institutions to Linux. Duncan said: "Backup has always been an annoying problem." He pointed out part of the IBM's Tivoli Storage Manager, which has been using the hot backup agent on the Oracle database running on AIX and Windows NT, but this agent It is still not available for Linux. Duncan also found that the server may have problems with SAN support.
The cost of safe management seems to have always been a management expense area of Linux than Windows. Duncan said that an important reason for Cedars-sinai migrates to Linux is due to hot repairs, service packs, etc., "We need to make a lot of changes in the NT server, we spend more time to modify the time on the NT server, we You must be able to set up a server, then ignore it, it is from this point of view, we migrate to Linux. "
Support fee
Linux support is said that Linux has a lower support cost, and there are more vendors to support it. They call Linux machines often run a few years without restarting.
When a new Linux machine crashes, replace this machine, because the cost of replacement is cheaper than repairs, either uses a method called "Google Service", which is to find a solution on the web newsgroup and message board. Enterprise users are gradually eliminating their doubts about this support method. Previously, Linux support was an obstacle to hinder the use of enterprises, and now it is becoming a strength.
Users may choose from hardware or software vendors, distributors or third parties, which may appear on Windows, but when using the Microsoft solution, the provider cannot actually modify Windows, and all Linux Support Vendors It also has this capability.
Want to migrate? Consider: Linux begins to enter the company, and technicians can assess when to migrate is reasonable. What is the biggest winner OS purchase fee of Linux? What is the best winner OS purchase fee? What is the cost of all software stacks on the basis? Rapidly developed companies will benefit from lower unit machine purchases. Hardware purchase fees can save money by using inexpensive Intel clusters, and avoid "hardware locking"? Currently using UNIX institutions, rapidly developing enterprises and horizontal expansion environments. Support cost Key ISV is as strong as Linux support as they support other platforms? Can you use "Google Services"? Enterprises with less complicated support or more internal Linux professional knowledge. Does the system management fee linux have the key tool you need? Can you benefit from management experience when you expand? Have an university that masters Linux system and uses UNIX, and a company with a variety of OS platforms that can be integrated. What is the security cost and risk costs for each platform implied (risk and patch management resource cost)? Enterprises that give up often repaired Windows environment. Is a major migration of disposable migration and conversion costs inevitable? What is the cost of re-training? Institutions that have been run on the application server that facilitates migration. What is the cost of integrating UNIX, Windows or legacy applications to a new Linux environment? UNIX applications are larger than Windows applications. Linux has been widely deployed into numerous data centers, usually used as a web server or file server, can also be used to handle network tasks such as DNS and DHCP, but used as a platform for running key enterprise applications, Linux does not seem to be optimistic, Usually used in this occasion of Solaris, AIX or HP / UX, providing the application with the highest level of performance and extension capabilities. However, many people think that with the release of the Linux V2.6 new version of the core, this situation will change, that is, Linux V2.6 new core will become a powerful promotion of Linux into enterprise-level critical applications. So what did the new version of the V2.6 have brought?
Next article
Larger, stronger kernel Linux V2.6
The Linux V2.6 kernel created a new era that supports large-scale operations and larger loads, allowing Linux to compete for high difficult tasks currently performed by Solaris, AIX or HP / UX.
The most important feature of the V2.6 kernel is to support large server architectures. It supports maximum 64GB of memory, which can handle more than 2TB file systems, and support 64 CPUs in X86-based SMP systems, which make the kernel and Linux more eligible to become a key task system running platform. The new kernel also supports NUMA (non-uniform memory access) system, the next generation SMP architecture, and PAE (physical address extension). The kernel breaks some of the people who are strong on the Linux header from the beginning, the number of main equipment has reached 4096, and the secondary equipment number of each primary device can be as high as 1 million, for most users, These numbers are far more than the actual needs, but for those enterprise systems that need to deal with many devices, this solution is a huge leap. In addition, the new V2.6 kernel also contains NPTL (local POSIX thread library) for replacing V2.4's LinuxThreads, NPTL enables Linux to have an enterprise-class thread support, which makes far beyond LinuxThreads. Level.
Born speed
Prior to the V2.6 kernel release, the way Linux performs tasks is the first to first serve, which interferes with the interrupt task of the kernel to handle additional processes or functions. However, the V2.6 core can be prioritized when needed, and allocate resources to those who need to process immediately, and then continue the previously interrupted tasks. This interrupt will only occur in a second time, and usually will not be noticed, but it will feel that system performance is very smooth. The V2.6 kernel does not make Linux a real-time operating system, but it is working hard to ensure that the task is handled and completed at any time. The core of these enhancements is a new process scheduler, and the process scheduler in the kernel can allocate CPU resources between system processes. The performance of the scheduler directly affects the latency of the system's response and process. In the V2.6 kernel, the new 0 (1) scheduler integrates new algorithms, enhances the performance of the system, especially the performance of interactive tasks. V2.6 has two new I / O schedulers. Using a predictive program in a default, it can effectively improve the performance of I / O scheduling, ensuring that all processes can get the required I / O time, without queuing. The other is a time limit scheduling program that can be used in three queues to request allocation time, while the predictive scheduler only tries to predict the process I / O request before these process requests actually issued. Regardless of which methods are used, new processes and I / O schedulers have greatly surpassed V2.4 levels.
V2.6 kernel test
Using the environment where users most likely applied to test new kernels are very helpful to comprehensively understand the new kernel.
The United States "NetWork World" has done a comprehensive test, comparison 2.4 and 2.6 two kernel performance. Tested on three completely different hardware platforms: Intel Xeon (x86), Intel Anteng (IA-64) and AMD Opteron (x86_64). The basic operating system uses the Red Hat Linux Enterprise Server V3.0, but kernel tests need to be specifically compiled for each server. V2.4 test is the official V2.4.23 kernel, while V2.6 test is the official V2.6.0 core. In order to run on a different platform, the kernel must be targeted, for example, X86_64 patches for the AMD64 platform for X86-64.org.
File Sharing Test is specifically used to simulate standard Samba server loads based on Samba V3.0.1 with this ground verification. On the Xeon system, the average throughput rate of the V2.4 core is 38.85Mbps, while the V2.6 core reached 67.30Mbps, and the increase of the increase of 73%. There is also a similar performance difference between the two cores in the Angar test, and V2.6 has a leading range of 52%. On the Opteron system, the V2.4 core rate is 49.37 Mbps, and the result of V2.6 is as high as 72.92Mbps, and the increase of 48% increase.
Performance improvements in Samba tests should be attributed to improved scheduling programs and I / O subsystems in the V2.6 core.
The database test environment is based on MySQL V3.23.58 and runs on the SQL benchmark set provided by MySQL. On the whole, the V2.6 kernel has completely exceeds the V2.4 core in the performance of the database test, especially in the Angang server, compared with the V2.4 kernel, the performance improvement of the new version of the kernel reached 23 % (Lead 519 seconds). On the Xeon platform, V2.6 exhibits 13% upgrading (leading 200 seconds). On the Opteron server, the speed increased by 29% (lead 415 seconds). In all tests, the most impressive thing is that the table insertion, the performance of the V2.6 core is increased by 10% (lead 100 seconds) on the Xeon server, while the performance on the Opteron and Angar platform is more powerful. Web server test also shows a very huge improvement. The static page test uses an HTML page with a size of 21.5kb, with two 25KB images from Apache 2.0.48. The target of the test is to use the APACHE AB reference tool to measure the number of requests per second. On the Xeon platform, the V2.6 kernel is less than 1000 requests / seconds than V.2.4, and the increase of 40% increased. On the Angang server, V2.6 has an increase in performance by 47%.
The web application test uses a custom CGI script written with Perl, pointing to the MySQL database running on the same system. The overall number shows that the result is increased than the static test, but the Opteron test is exception, and the general performance improvement in all platforms is between 14% and 22%.
Direction of development
Experts pointed out that IT companies running large databases and other key task applications on the V2.4 kernel should begin a preliminary test of V2.6. V2.6 kernel is the future leader, and companies should begin familiar with this new kernel capacity and do some preparation for future applications.
In order to make full use of the V2.6 kernel's enhancement, the administrator needs to configure and establish custom kernels for their specific workloads when building server architectures, but the problem of such specific use cores is that they exist with the officially released core Many differences, whether in the default option or the patch contains different. On the other hand, these kernels typically support hardware is very wide because almost every module has been configured, and the hardware compatibility of the target system can be ensured.
At the same time, there are many defects in the V2.6 core, and kernel developers are also actively looking and resolve these issues. If you start considering porting to the V2.6 core, it is highly recommended to fully test the new kernel before any productive implementation is performed.
The V2.6 kernel is proud to have many new features and significantly improve performance. The kernel will also make Linux ready to meet a new phase of the challenge: real enterprise-class critical applications. In order to preempt the data center market, Linux must be able to adapt to the needs of the existing users at any time, and continuously expand new areas, pull more users into the Linux camp. From the current situation, the V2.6 kernel has been prepared for this task.
