Others

Marketing in China

The great myth propounded since the opening up of China was “…If the 1.3 billion mainland Chinese buy only one of your products, you will make a huge fortune…”. According to Professor Wang Zi-le of China’s Institute of Research on International Trade and Economic Co-operations, only approximately one-third of the 394,000 foreign-invested enterprises in China returning a profit in 2002. Only three products have been bought by almost each of the entire Chinese population – Coke, Pepsi and Wahaha.

According to Professor Wang, the mere reality of the large influx of competing foreign firms itself posed the biggest obstacle for each to realize their anticipated profits. Local protectionism and bureaucracy also contributed to the nightmares of many. Those foreign investors who came to China with outdated products and technology or a second- class management team were pressed the hardest. “…Those multinationals came in with the sweet dream of easy dominance of China market, and thus huge profits said Wang. …Once they got their body and soul inside, they began to feel the blazing heat and numbing chills of this market…”

One Country – Two (Market) Systems

China never was “a market”, says Yao Xiaoqian from IBM’s Management Consulting Department. He adds that it is difficult for any foreigner to imagine the huge gap between the cities and the rural communities in China. While internationally the income gap between urban and farming populations is approximately 2:3, in China the ratio gap is well over 1:2.8. Because of this, the urban and farming population’s makeup two distinctly different markets exist in China. If companies persist in maintaining the same brand name and product lines for both markets, they find it necessary to cut prices drastically before the products that are popularly sold in the cities become affordable or acceptable to the rural farming community markets.

Before deciding to go for broke with the one-market one-brand ideal, managers may benefit by taking another look at the map, the logistics, and the costs to the supply chain.

China is about 5000 km from East to West and more than 4000 km from North to South. Temperature differences between North and South can reach as much as 50 degrees Celsius. The rugged terrain in many parts of China that time and time again even bogged down the imperial armies remain a significant test for MNCs. Quite sensibly, most MNCs limit their marketing thrusts to the urban markets, leaving the rural market to the locals.

The Big Organization Disease

Many MNCs in China have also become infected by the “Big Organization Disease” – the same disease that either fatally or incurably infected many of China’s own state enterprises or characterized by the rapid growth of the organization particularly in middle management levels. I have previously referred to the San Zhu Health Potion case, where an organization grew to a staff complement of 150,000 within 3 years and proceeded to bankruptcy within 7 years. Prior to Unilever China’s vigorous restructuring in 2001-2002, its expatriate middle to senior management complement grew to approximately 120, compared to 30 after their “forced diet”. MNCs should watch their “diet” and not get too carried away by the dream of catching this huge dragon in a hurry. The mere cost of management talent, especially expat talent, is a key handicapping factor to newly arrived MNCs in a market where they face grassroots locals long experienced in price-cutting wars and marketing guerrilla warfare.

The Fractured Organization Disease

During the first twenty years of its opening up, manufacturing and marketing in China was localized or regionalized. Market players were mostly state-owned and heavily protected by their state, provincial and local governments. Many early foreign entrants into China either sought or were guided by the state towards joint ventures with these state enterprises. Once a joint venture was established, it was also customary for the local government to send in a local management team to provide a “balance” for foreign management. For example, as soon as Beijing Dafa Zhengda was established between the C.P. Group of Thailand, and their local partner, the Chinese side sent in 6 deputy general managers who previously held government positions of section chiefs. At the Zhenda Ningbo Company, another C.P. Group joint venture, the Chinese side also sent in 6 former government officials, one of them the “native king” (refer to earlier article) from the era of Cultural Revolution. It was an open secret that these local deputies were there to ensure the maximum benefit for the Chinese side. Amongst the tasks they were most interested in controlling was construction project bidding and controlling, procurement, and personnel recruitment, selection and management. Thus, when the C.P. Group tried to consolidate raw material procurement of its 100 feed mills in China, it encountered huge resistance from the local side. Such was the same scenario in Unilever which had more than a dozen joint ventures in China. In order to ensure cost control, Unilever sent in more and more expat management until they finally realized that there was more to cost control in China than devising policy and accounting.

The difference between the C.P. Group and Unilever was that all C.P.’s 100 feed mills produced similar kind of feeds and the management of the individual joint ventures was forced to compete with one another in cost effectiveness and profitability. The joint ventures of Unilever were making different products and could not readily be compared to each other. In addition, the local side could even be a rival or in diffident parties which were neither interested in c-operating with other Unilever joint ventures nor caring about anything other than benefits to the individual local side itself. Thus, when Unilever attempted to buy out all the joint venture into its holding companies, it encountered strong opposition from its local partners. Unilever went ahead anyway and proved that it was all worthwhile doing.

The Proud Conquerors’ Disease

China’s ancient book of Warfare read that “The proud armies are bound to be defeated.”

When MNCs first entered the China market it was 50 to 100 years behind the West. They came with high confidence and looked far into their “New World” with golden dreams although being troubled occasionally from gaps of cultural understanding and the divergent governmental administrative systems. With their vast capital, technology, and product lines, early arrivals such as Coke, Pepsi, and P&G practically scooped up their entire target markets with the power of their marketing offenses and buying out of existing local leading brands. Having gained overwhelming dominant market shares with relative ease during their early years in China, they might even have considered simply gliding along happily ever after.

New Giants Rise from the Grassroots

During the mid 80’s to mid 90s, it was practically unthinkable to imagine any real competition arising from within China itself, especially from previously little known manufacturers. But many MNCs were rather unceremoniously given a rude awakening when reports from major marketing research companies such as A.C. Neilson began to show companies like P&G, Unilever, Cisco, Henkel and IKEA that they cheese was no longer as readily available. Local products had either overtaken them or already matched their market-share.

The Countryside Surrounding the Cities

Very possibly the only reason that provincially targeted companies like Nan Feng, Si-bao and Zhejiang Nice were not bought up, squeezed out, or smashed up by P&G and Unilever was that they were too small, too remote, and too insignificant. Early on they were not within the competitive sights of the giant MNCs.

Nan Feng positioned itself as the maker of low price detergent for rural consumers. It had an advantage of being situated on the shore of a large salt lake in the countryside of the remote Shanxi province. Readily available raw materials helped them succeed with a low cost low price strategy. In 1995, it started driving its market by painting 600,000 square meters of crude advertising for “Qiqiang” on brick walls and smoke stacks mainly in the northern part of the country (being furthest from the P&G territory which produced its products in Guangzhou in the south). Its first sales promotional team consists of 3000 sales persons being supported by the “drums” and “gongs” band in the north and the lion dance troupes in the south. All these ads and campaigns were considered rather primitive and crude in the eyes of most modern consumer product managers. As the local detergent factories became squeezed to the verge of bankruptcy by P&G during the early 90s, Nan Feng quietly merged them in and set up manufacturing bases in Xi’an, An-Qing, Guizhou, and Inner Mongolia, again areas not within focus of P&G’s strategists. By 1999, “Qiqiang” became the No. 1 detergent brand in China on the back of its mainly rural sales. Then in this very same year, it announced its invasion of the cities, beginning with Beijing and Shanghai.

Zhejiang Nice was ranked the second smallest detergent factory in China, situated in the very remote town of Lishui near the border with Fujian province. So behind the outside world was Lishui, that the inhabitants of this town raced to see the first train passing through the town in 1998. Many in this town had never seen a train in their lives. By the year 2000, it was hauling in 340 million yuan of profits from sales of 2.5 billion. Its profitability was nearly 25 times that of the next competitor.

Zhejiang Nice started manufacturing “Diao Brand” detergent in 1999. By 2000, it’s own and over two dozen contracted factories simply could not satisfy the rapid rise in consumer demand for “Diao” products. Interestingly, four of the OEM factories belonged to Henkel of Germany and another two belonged to Procter and Gamble. The 29 OEM factories spread across 19 provinces in all parts of China.

The OEM localized manufacturing saved Nice approximately 600 yuan RMB per ton of detergent produced in inward and outbound transportation costs. After paying 200 to 300 per ton to its contracted manufacturers, Nice could very well afford its low low price strategy and still reap a huge profit from its sales of 500,000 tons in 2000. With its OEM job orders, it helped the Xuzhou Henkel factory recover from its 40 million RMB losses and make it profitable. Nice expected to manufacture and sell more than 1 million tons in 2001. In 2001, Zhejiang accounted for 95% of profits made by all firms in detergent industry in China.

Si-bao was a manufacturer of skin-care products based in Wuhan. Having carefully market tested its hair shampoo products, Si-bao decided to give its shampoo a brand name of its own, instead of using the same brand name as its skin care product. So “Slek” was born. From the very beginning, it foregoes the then widely used nationwide network of distributors. It started building a marketing network of its own. Through massive advertising in its home base city of Wuhan with the appeal “Double action in-depth protection while giving the natural bright beauty to the hair”. The tag line proved to be an overwhelming success. In 1997, Slek set up branch companies in 12 provinces with 63 marketing and sales centers employing more than 20,000 staff, mostly POP sales staffs. Even at this stage, its impact may well still not have been of much interest to P&G.

In 1997, the Slek cal very introduced its shampoos into tens of thousands of selling points, putting up POP banners and sign boards that turned many stores and their own in-store sales promotion centers into seas of red. The instruction from HQ was to take control of all of the most prominent point of purchase displays, and to be seen side- by-side with P&G shampoos. In 1998 as P&G focused attention on delivering a knock-out blow to local competitor “Olive”, it failed to pay any attention to the advancing “Slek”, thus helping “Slek” gain market share to the extent that it surpassed the former No. 4 brand – Unliver’s “Lux”. Within the very next year, in 1999, it overtook Pantene’s No. 3 spot with 10% market share compared to Pantene’s 6.5%, becoming one of the top 3 brands. Its market share continued to rise to 15% in 2000.

Who Moved My Cheese?

According to A.C.Neilson, the two top selling detergents in China market in 2002 were “Diao” (Vulture) and “Qiqiang” (Surprisingly Strong) – both new local brands from Chinese manufacturers. They were followed by “OMO” and “Breeze” from Unilever. Proctor and Gamble’s brands just made the last two spots on the Top Ten list.

2002 Market surveys report the top selling hair shampoo brands as Rejoice, Slek, and Head and Shoulder. Pantene, another proud brand of P&G had already dropped from the top three brands since 1999. While in 1997 P&G was still mustering nearly 70% of the market (Rejoice – 35%, Head & Shoulder – 17% and Pantene – 16%), by 1999 Pantene was assaulted by a proliferation of counterfeit products and their market share dropped to 6.5%. Slek took over their third position with 10% market share. With an extremely aggressive point-of-purchase sales promotion onslaught, Slek’s market share rose to 15% in 2000 replacing Head & Shoulder as No. 2 behind Rejoice with a dwindling 30% market share. The preliminary figures for 2002 must certainly be disconcerting for P&G and Unilever, as combined they now command only 55% of the Chinese hair shampoo market. P&G’s CEO is philosophical – stating recently that – “Slides in market share is expected after our early successes in dominating the market”.

P&G and Unilever are not alone in experiencing a rude awakening to this new China market. The world’s leading carbonated soft drink makers, Coke and Pepsi either bought up or squeezed out all 8 major beverage manufacturers that existed during the early 80’s. Again hidden warriors emerged from the carnage, in the form of previously unknown and untested new competitors Wahaha and Kangsifu. Wahaha started with a 140,000 yuan loan and 3 employers as a small factory in a school distributing stationery and ice cream in 1987. In 2001, it sold 2.5 million tons of bottled water, milk and cola type beverages attracting a total income of 6.2 billion yuan. During the same period Coke sold 1.3 million tons and Pepsi sold 900,000 tones of their cola-flavored carbonated beverages. The instant noodle champion Kangsifu won 47% of the approximately 300 million ton bottled Chinese tea drink market during the same year.

The Company Still Makes the Product, but the Brand Doesn’t Sell Them Any More

According to Chu Ke-xiang from IBM Management Consulting – “Amidst this on-going globalization and WTO trend, Chinese consumers are moving away from “name brand” or “foreign brand” preferences… …Where foreign named brands used to bring trust, reliability and brand loyalty, Chinese consumers seem to have grown out of that now.” Research indicates that Chinese consumers now spend more time in front of display shelves, reading product descriptions with an interests never exhibited by consumers before. They ask point of purchase sales staff pointed questions. In all, it is clear that the brand name is no longer the almost sole factor in Chinese consumers’ buying decisions.

Price Wars and Anti-Dumping Tactics

In the microwave oven market, Galant triggered the first price war in August 1996 with average price reductions of 40%. Fourteen months later it announced another price cut across the board of 29 to 40% and in 1998, it launched a huge give away “Buy 1 and get 3 free” campaign increasing its sales volume to 4.5 million units. They soon dominated 60% of the China market, and practically pushed all foreign brands to the side lines. This should have satisfied the company’s desire to become the No. 1 microwave supplier in China and in the world. Not so. In June, 2000, Galant introduced another 40% across-the-board price cut aiming at totally eliminating their rivals. Four months later, in October 2000, it cut its prices for all hi-end microwave ovens by another 40%. Again, in January, 2002, it lowered the hi-end prices by another 30% in order to “make the hi-end lines accessible to common Chinese consumers”. As if being the world’s largest microwave makers was not enough, in February, 2002, it cut the average prices of its air conditioners by 35%.

If Galant sold its product overseas at prices lower than its domestic prices, the FTC and the EEC would certainly huddle together and deliver an “Anti-dumping tax” on the feisty company. But what can they do if it reduced the prices in its own home market?

In an earlier article, you might recall a curious promotion of selling color TVs by weight. If so you will remember that this was a “sarcastic” move to protest the reduction of color TV prices to an unthinkable level – below $50 USD for a 21 inch new color TV and $150 USD for 29 inch models. In this market of approximately 20 to 21 million sets per year, Changhong sold about 8 million sets, followed closely by TCL, Konka, and other local brands. A visit to any store nowadays will confirm that foreign brands have almost disappeared from the under 29-inch TV display sections. The ongoing price war certainly hurt local makers as aggregated profits in the past years amounted to only 540 million comparing to Sony which made 1 billion yuan profit on sales of merely 1 million sets. Toshiba sold only half a million sets and made about 500 million yuan in profits. These Japanese firms have discovered for themselves some Anti-Dumping protection that works more effectively than the FTC and EEC!

Car Racing, Forced Marriages and Bigamous Partner

As mentioned in earlier article, Volkswagen boasts one of the most successful sino-foreign joint ventures (with Shanghai Automobile), their Santana model becoming almost synonymous with automobiles in China. However, when First Automobile of Changchun was selected to join Volkswagen in the latter’s second joint venture in China in 1991, the partnership did not prove as successful. By 1996, when the Shanghai venture had reached 200,000 units in car sales, the Changchun venture had provided less than 10,000 units. Rumours that the German side was considering withdrawal from the joint venture were frequent.

In 1996, a new general manager from the Chinese side, a Mr. Lu Lin-nai who speaks fluent German was appointed. Sales starting picking up immediately with favorable recognition for the reliability of its Jeda model immediately. In 2001 Jeda sales surpassed 100,000 units and then 200,000 units in 2002. As of the end of 2002 back log orders for the car have already taken care of production for four months ahead. There had been suggestions that production capacity would be expanded from its current 150,000 units per year design with an investment of 2 billion yuan, but the expansion plan was flatly rejected by the Chinese side. The Chinese partner preferred receiving its share of profits in dividends than to continue with re-investment in the joint venture

“Why make more Jeda’s, when the brand belongs to VW anyway?” was the comment overhead from one insider and was quite representative of the stormy relationship between the partners on-going on from the time of sickness to the time of good health.

Festivities to celebrate the sales success in Changchun were abruptedly cancelled three days before the scheduled date of December 23, 2002 due to “Special circumstances.” It was reported that key top executives of First Automobiles were busy with other engagements elsewhere and were not available for the celebration. Mr. Lu’s 4 year term was extended by two years in 1999 on account of his outstanding performance in turning the company around from the red but Lu left his job as the general manager of the joint venture at the end of his extended term in 2001. He was stripped of his political party membership, “detached” from the company, and rewarded with the dismissal of all of his immediate relatives in the company’s employ. He was accused of siding with the foreign partner during his tenure while he failed to satisfy the local side’s desire to take out their profits rather than reinvesting the average annual profit of 1 billion yuan each year back into the company. He was also accused of having not done enough to obtain the German side’s support for the upgrading of the local side’s own product, the Red Flag. The German partner had provided the partner with the technology from its 1980 Audi model as part of the joint venture agreement and had since hesitated to help upgrade the Red Flag, – the pride of local Chinese automobile industry. It also charged a 15% management fee on spare part purchases from Germany.

On the date scheduled for the celebration of First Automobile-Volkswagen’s Jeda sales, the new general manager Mr. Qin Huan-min was not in Changchun but in Tianjin – joining the celebration with its other partner Toyota Motors to mark the first jointly produced Toyota Bora from the production line.

The joint venture agreement signed in August 2002 called for an investment of 23 billion yuan to produce 300,000 to 400,000 units of luxury-to-compact automobiles annually. The Japanese side had agreed to provide First Automobile with technology from Toyota to raise the Red Flag to a brand new level. Toyota had also agreed not to charge any management fees on purchases of spare parts for the jointly produced automobiles.

For Better or Worse, Understanding of the Locals Counts

After they have skimmed through their Chinese economic and marketing data and information, MNC strategists are well advised to take a good look at Chinese historical stories where more often than not, “A small flickering proceeds to consume the entire mountain and plains”. Even the ruling Manchus found themselves naturalized by the Han culture 300 yers after their reign over the Han people.

As one Japanese executive – one of thousands being sent to China from Japan – recently told me in his perfect Mandarin:

“…We understand Chinese a lot better than those Westerners. That’s why we have more successes than the American and the European…”.

Source: P. Wattanavitukul of P. W. Consultants specializing in Investment, Management and Trade in and with China and Human Resource Development in Shanghai and Ningbo.

Monday, March 1st, 2010 Others No Comments

This tip for SEO Programming

Understanding the criteria that influence search engine rankings
Building keyword-rich URLs using PHP, Apache, and mod_rewrite
Using the HTTP headers to properly indicate the status of web documents
Relocating content using the proper HTTP status codes
Coping with duplicate content effectively to avoid penalization
Avoiding being the victim of black hat SEO techniques
Implementing geo-targeting and cloaking using IP delivery
Promoting your site using web feeds and social bookmarking
Optimizing your web pages using better HTML and JavaScript
Enhancing existing web sites
Using search engine and traditional site maps effectively
SEO for WordPress blogs
Building a SE-optimized E-Commerce catalog case-study

Friday, October 24th, 2008 Others No Comments

Mover谈网站标题优化

一个网站的标题优化是最简单的，这是与网站其它方面优化相比而论。

对于各种搜索引擎来说，一个网站的标题对确定内容的相关度具有决定性的作用。因此，写好一个好的标题，也是需要注意一些问题的。

1、关键字标题。对于这种标题来说，标题中除了关键字，没有别的内容。这种优先度是最高的。

2、关键字只构成标题的一部分。Google和百度通常认为关键字在标题中越靠前，那么它的优先性就越强。所以，当我们要强调某个关键字的时候，一定要尽量将它往前靠。不过，未必排在第一位的关键字就是最优先的。因为Google要考虑它的在标题内容的占有的比率。这种比率并不是越高越好，而是有一定限制的。最佳比率不能超过20%，但这仅仅是作为参考的比例。原因是，当Google数据库里本身具有这样相同的标题，Google会把发表时间的先后、网站本身已存在的优先程度作为考虑因素。这里不作详细说明。

3、关键字的分割与连接。如果希望在相同的页面里面做多个关键字的优化，那么将要面临关键字的连接问题。我们经常看到有许多的网站使用空格、”_”、”-”、”|”。那么到底哪种连接符是最佳的呢？我建议英文用”-”，中文用”|”或”—”原因是什么呢？因为Google把”|”连接的中文字作为一个关键字标题，这样就不存在关键字比率的问题了。但是还是存在分词的比例问题。所以我建议一个页面做一两个关键字的标题是最好的。最好不要用连接符堆砌关键字。Google目前并没有限制一个标题的长度，但是你如果希望你的标题更友好的话，中文字最好不要超过30个。其它搜索引擎的标题长度也有一定限制，大家可以注意去观察。

Friday, October 24th, 2008 Others No Comments

压力测试

不是专职做压力测试这行当的，只能是以自己的经验来以外行人的眼光来说说压力测试，压力测试并不仅仅是个压力测试的过程，而是一个相当系统和复杂的工程，我认为压力测试是为了让系统达到所期望的运行效果以及承受所期望的压力，这也就要求压力测试应该帮助性能调优团队，为其提供一定程度的指导，在这里我不将压力测试和性能调优分的那么清楚了，在我看来，压力测试过程包括了：明确压力测试的目标、制定压力测试方案、进行压力测试、分析压力测试结果、寻找瓶颈并进行调优以达到目标，在这篇blog中来细看下这几个过程以及常用的方法。
不是专职做压力测试这行当的，只能是以自己的经验来以外行人的眼光来说说压力测试，压力测试并不仅仅是个压力测试的过程，而是一个相当系统和复杂的工程，我认为压力测试是为了让系统达到所期望的运行效果以及承受所期望的压力，这也就要求压力测试应该帮助性能调优团队，为其提供一定程度的指导，在这里我不将压力测试和性能调优分的那么清楚了，在我看来，压力测试过程包括了：明确压力测试的目标、制定压力测试方案、进行压力测试、分析压力测试结果、寻找瓶颈并进行调优以达到目标，在这篇blog中来细看下这几个过程以及常用的方法。
明确压力测试的目标
通常来说（注意是通常），压力测试的目标有这么几点：
1、评测系统是否满足压力支撑的要求
要评测系统是否满足压力支撑的要求，同样要做的就是需要明确定义系统需要支撑多大的压力，例如：
机器的配置：CPU、内存、硬盘、etc.
网络条件：100M
操作系统：Linux core: 2.6
当并发数为10用户时，系统应能在20ms内响应完毕（这个时候的TPS为500），系统的load需在2以下；当并发数为100用户时，系统应能在 50ms内响应完毕（这个时候的TPS为2000），系统的load需在4以下；当并发数为200用户时，系统应能在80ms内响应完毕（这个时候的 TPS为2500），允许其中有千分之一的出错率，系统的load需在6以下，在压力测试的过程中，只要其中的任何指标未达到，均可判定系统尚未达到压力的目标。
实际的压力测试的这个指标会比我这里举的例子复杂很多，例如还需要考虑网络流量、内存消耗、IOPS、连接数等等。
这里面压力测试隐藏的目标是为容量规划提供一定的指导，例如目前的系统在某种配置的情况下单台机器能承受的最大并发数为100用户，那么如果系统的高峰压力是1000的话，如果系统支撑无损水平扩展的话就意味着需要10台这类配置的机器，这一步同样是经过测试的。
2、预估系统上线运行的状况
毕竟通常压力测试环境和线上的环境是会有很大的不同的，压力、数据量、硬件环境等，基本上只能是根据线下的环境的情况进行一定比例的对比后计算来预估，这里面很重要的是要预估系统上线后正常情况下的表现状况、一定的增长比率后的运行状况以及风险点（例如当并发用户数增长到多少时、系统load到多少时可能会出现问题）。
这一个目标我觉得非常难达到，但随着经验，我相信是可以做到的，如果能做到这种效果的话是会有很大的帮助的。
以上这个两个目标基本是压力测试都要达到或希望达到的，而具体有可能会因为系统的业务的具体情况会制定其他不同的目标。
制定压力测试方案
这步是压力测试整个过程中最难的步骤之一，为了能够测试出系统是否符合压力支撑的要求以及评估上线的表现，通常我们会希望搭建出和生产环境完全相同的环境，但这就是最麻烦的一点了，基本上是不太可能的，因此通常能采取的方法会是：
1、做等比例的缩放
按照生产环境的情况做一定比例的缩放，例如生产环境的数据量为1亿条，那么测试环境等比缩放到5000w条，生产环境的处理速度的情况…；
2、更差环境、更高压力的测试
采取比生产环境更差的机器配置、网络环境来进行测试，例如ebay的要求是能够承受10x的压力。
3、仿真测试
据资深人士而言，仿真测试要做到基本是不太可能的，仿真测试首先要求的是收集到生产环境中的运行状态的数据，然后在测试环境中编写程序来做到模拟生产环境运行的效果，这个难度基本是非常高的。
我自己现在做压力测试更多采取的做法是以上三种方法的合集。
在确定了测试方法后，就基本可以确定压力测试的环境了，环境确定好后需要做的是压力测试的案例或场景了，在压力测试的案例中需要涵盖正常场景以及异常的场景，正常场景是非常容易做出来的，只是需要根据生产环境收集的数据（例如不同级别的用户比例通常是7:2:1）或预估的数据来搭建相应的测试案例，异常场景则是比较复杂的，需要考虑很多的因素，例如数据库出现异常、网络出现异常等，这里我觉得通常的做法是画出正常场景的处理流程，然后划分交互边界的信任边界，对于所有的第三方交互都认为是不可信任，例如不能信任调用数据库是一定会快的，或一定会成功的，在异常场景中应涵盖这些信任边界的异常状况的测试，这些对于高可用性的系统而言是非常重要的，几个9的成败就在此了，^_^，当然，高可用性又是个更复杂的话题，不在这里讲。
在压力测试方案中还需确定的是考评的指标，通过会包括：tps、系统load等等。
进行压力测试
相对来讲，在有了压力测试方案后，这一步并不是什么太复杂的事情，只是需要选择一个和压力测试方案比较符合的工具来执行，例如jmeter、 loadrunner等，当然，这些工具相对来说也是比较复杂的，而且之间的差距也是比较大的，至少目前来看，jmeter和loadrunner的差距还是不小的，尤其是需要进行高压力的测试时。
分析压力测试结果
这步同样是压力测试中很难的一步，在这一步需要做出的根据压力测试的结果分析出系统的具体表现情况，判定系统是否能够满足压力指标。
以loadrunner产生的分析结果图而言，通常需要分析以下图：
1、Total Transactions per Second
这张图中显示了系统在进行压力测试过程的TPS的变化情况，从这张图中我们可以看到系统的TPS的情况，通常来讲，随着用户数的增加，TPS应该是呈一定比率的增长的，等增长到一定程度后会达到瓶颈，甚至开始下降，这也是TPS的瓶颈值了，这张图可以帮助我们评估系统的TPS是否符合要求。
另外，在这张图中，我们可以看到系统从什么时候开始出现出错的transactions，从而判断出错率是否在可接受的范围。
2、Transaction Response Time Under Load
这张图非常的重要，借助这张图我们可以分析随着用户数增加的情况下，系统的劣化状况，最佳状况当然是一条直线，但这基本是不可能的，毕竟资源是有限的，需要判断的是劣化的程度是否为可接受范畴。
另外就是需要关注数据中90%的用户的响应时间的状况，如果少量用户响应慢是可接受的话，那么有可能在之上指标不达标的情况下仍然满足了压力指标。
3、Unix Resources
这张系统load图自然是非常的重要，借助这张图大致可以判断系统随着用户数的增长消耗的资源的变化情况，这对于调优以及容量规划而言是很重要的，但还是得取决于应用本身，:)。
loadrunner还提供了其他方面很多的图，可以根据考评的要求来自行进行分析。
寻找瓶颈并进行调优以达到目标
这步不属于压力测试范畴，但还是在这里稍微讲讲，毕竟压力测试结束后如果系统没达标的话就必须进行这步了。
寻找瓶颈，这自然是非常难的事了，通常系统达不到要求的状况都会是随着用户的增长，响应时间劣化的过于厉害，在这样的情况下首先得观察系统硬件资源的变化情况，如果是硬件资源耗尽的话，需要查查为什么资源被耗尽，假设最后判断确实需要耗费这么多的硬件资源的话，也许需要考虑增加硬件资源或是水平扩展，否则的话可能需要从软件层面相应的优化系统了，这样的话可以进入下一步了。
如果不是硬件资源的限制的话，得在系统中从头到尾设置时间跟踪filter，从而判断响应时间劣化的原因，看看是系统中哪些步骤造成的，这个是细致活了，有可能要查非常久。
其实这里说的还是相当的简单了，在寻找瓶颈的过程中是个非常繁琐的过程，需要不断的尝试，硬件的增加、OS的调优、jvm的调优以及软件系统本身的调优等等，这些很多时候需要的是经验，因此某知名人士曾经说过如何寻找瓶颈和调优，其中依靠的一点就是直觉，^_^。
当然，在寻找瓶颈的过程中，可以借助os的工具、java的工具（例如gc打印、jprofiler等）来进行查找。
（ps: 不过感觉很多情况下都是应用本身造成的性能瓶颈，在写程序时稍不注意用错一个数据结构都有可能会导致比较大的问题，所以我现在查找瓶颈的时候更多的还是先从软件本身下手，只是软件性能要做到提升通常来付出较大的代价，这个时候需要权衡）
调优基本上要求对硬件、OS、JDK、数据库甚至软件的实现方式等都要有非常深入的理解，至少要能做到判断出瓶颈因素，然后找相应领域的专家来解决，因此要求是非常高的。
关于性能调优的知识体系这里有篇不错的文章：

http://www.cnblogs.com/jackei/archive/2008/06/27/1231307.html

话题太大了，写到最后发现基本上还是有些泛泛而谈了.

Tuesday, September 23rd, 2008 Others No Comments

S600 硬键盘无法输入的解决办法

现在智能手机越来越受大家欢迎，但是应用服务还是一个问题。今天有个朋友让我帮忙解决一下手机键盘无法输入字母及汉字的问题。我猜想应该是需要安装一种输入法。果然很快找到的解决方案。特共享给大家使用。

下载：
plumsip6-pro-ppc10

Tuesday, September 16th, 2008 Others No Comments

Google Chrome 浏览器已发布

Google Chrome 浏览器背后的故事

Watch a video from the development team on the thinking and features behind Google Chrome.

观看一个来自Google Chrome浏览器背后的开发团队在这个浏览器的思考和特点的视频

Watch the video

观看视频

阅读这项技术

Look under the hood of Google Chrome in this comics interpretation of key engineering decisions, by Scott McCloud.

请看下面连环画中对Google Chrome 浏览器的关键技术说明,由Scott McCloud而作.

Read the book

阅读

Download Google Chrome

下载Google Chrome浏览器

Wednesday, September 3rd, 2008 Others, SEO No Comments

影响页面优化效果的十要素

要达到良好的网站优化效果，除了要保证高质量原创内容及建立链接提高Link Popularity外，对网页进行充分的优化更是必需的。

以下部分按其对优化效果影响度自大至小排列, 仅供参考：

Title 标签

在HTML语句中包含在之间的内容，用来概括页面的基本内容。在用户浏览网页时Title显示在浏览器窗口的上方，而在SERP页面中，Title更是决定用户是否点击该页面的主要因素。

详细：

* 网页Title的优化的原则
* 网页Title的优化的实施

内容中的关键词

网页文本中包含的用户搜索时输入的关键词，这方面的优化包括文本中关键词出现的频次、位置以及组合方式等。

详细：关键词密度(Keyword Density)
页面中使用的Headings、粗体、斜体等

Headings，通常意义上指段落标题，在HTML中以H1、H2、H3之类表示。

通过使用Headings、粗体等字体方面的变化给出概括页面或段落的概要或强调部分，这在可以在给访问者以较好的阅读体验外，还能够有效地帮助搜索引擎判断页面的重点内容。交替使用则效果更佳。
站内其他页面指来的连接

一般意义上，如果某个网页的内容很重要，那么，同一网站内肯定会有相当多的页面指向该网页，呵呵，很好理解吧？事实上，搜索引擎也是这么想的，一个内部连接充分的网页同样会被引擎认为重要。这方面的优化包括连接来源页面自身的内容、位置、链接使用的Anchor text等。

页面的外出链接

这方面的影响可能是正面的也可能是负面的，取决于链接的质量、与本页内容的相关度甚至网站之间的相关度等。

详细：

* 网站的出站链接
* 出站链接与入站链接的平衡

网页存在时间

每个网页对搜索引擎来说都有一个“出生时间”，即其被搜索引擎的Spider第一次爬行收录的时间。存在时间越长的网页，搜索引擎便认为其更权威，可信度更高。当然，这是在保证该网页不是“死”页面，没有被埋藏起来的前提下。

在URL中使用关键字

在URL中使用的关键词，对其的更详细讨论可参见初步优化: 子域名、URL、Keyword与description、Keyword in URL : 中文网站的两难等。

网页大小

网页大小可以用文件大小KB来表示，也可以用字数来表示。一般说来，一个过大的网页（KB）会影响浏览速度，一个过长（字数）的网页则难以在关键词密度方面找到平衡，但具体很难有定论。

Alt与Title标签

这包括图片的alt、Tltle标签，也包括链接中使用的Title标签。虽然这常被SEO Spam用做关键词填充，不过，各大主要搜索引擎仍然使用其做参照。

HTML代码

W3C认证被许多网站用以证明其页面文件符合HTML规范，简单说来，一个HTML代码错误百出的页面对搜索引擎Spider的总计也会带来困难。但是否经过W3C认证就会有助于搜索引擎排名目前仍存在不少争议，甚至有的测试表明一个故意留上几个小错误（注意，是小错误）的网页甚至能更受到搜索引擎的青睐。因此，我们要做的倒不一定非得通过W3C认证，但至少要保证其在各种浏览器下显示正常，保证搜索引擎的Spider能够正常分析。

Tags: 搜索引擎优化, 网站优化

Tuesday, September 2nd, 2008 Others, SEO No Comments