网络文章@202312

这个FAQ内容比较多，退出云其实没有什么技术壁垒：硬件可以找承包商进行托管，自动化部署有现成的工具，大部分应用不需要有太多的弹性。唯一需要做的就是下定决心，和前期一次性的投入。

buy baseline, rent spike. 机器足够正常负载就行，当出现高峰的时候在去进行租赁，这样是最合适的，除非是每天都有spike.

Shortly thereafter, all the hardware we needed for our cloud exit arrived on pallets in our two geographically-dispersed data centers. All 4,000 vCPUs, 7,680GB of RAM, and 384TB of NVMe storage of it! 此后不久，我们退出云所需的所有硬件都装在托盘上抵达我们两个地理位置分散的数据中心。全部 4,000 个 vCPU、7,680GB RAM 和 384TB NVMe 存储！

但是，如何对服务器进行机架和堆叠以及拉动网络电缆呢？谁这样做？

We use a white-glove data center service provider called Deft. There are tons of other companies like them. And you pay them to unpack the boxes that arrive from Dell, or whoever you buy from, straight to the data center, then they stack it, rack it, and you see the IP address come online. Just like the cloud, even if it isn’t instant. 我们使用名为 Deft 的白手套数据中心服务提供商。像他们这样的其他公司还有很多。你付钱给他们，把从戴尔或任何你购买的产品送来的盒子拆开，直接送到数据中心，然后他们把它堆放起来，放在架子上，然后你就可以看到IP地址上线了。就像云一样，即使它不是即时的。

Our operations team basically never set foot in our data centers. They’re working remotely from all over the world. The operating experience is far more like that of the cloud than it is the early days of the internet when everyone drew their own cabling. 我们的运营团队基本上没有踏足过我们的数据中心。他们在世界各地远程工作。与互联网早期每个人都自己布线相比，操作体验更像是云。

需求爆发怎么办？自动缩放怎么样？ The shocking thing about buying your own hardware is realizing both how cheap and how powerful it’s become. The progress in the last 4-5 years alone has been immense. This is one of the reasons that much of the cloud is getting to be a worse deal by the year. Moore’s Law is hammering the prices and increasing the capabilities of stuff you buy from Dell and others. But it’s scarcely moving the needle on the price you have to pay Amazon and others for their managed services. 购买自己的硬件的令人震惊的事情是意识到它变得多么便宜和多么强大。仅过去 4-5 年的进展就非常巨大。这也是云计算业务逐年变得更糟糕的原因之一。摩尔定律正在降低你从戴尔和其他公司购买的产品的价格并提高其功能。但这几乎没有改变你必须向亚马逊和其他公司支付的托管服务的价格。

That’s all to say that you can afford to dramatically over provision your own hardware to give you amble headroom to deal with spikes, and it’ll barely make a difference in the long-run budget. 这就是说，您可以大幅超额配置自己的硬件，以便为您提供应对峰值的足够空间，而且这对长期预算几乎不会产生影响。

That said, if you regularly do face 5-10x or higher spikes in demand over baseline, you may indeed be a good candidate for cloud. This was after all the original incentive behind AWS. That Amazon needed way, way, WAY more performance on Black Friday or Cyber Monday than they did the rest of the year. So flexible hardware made sense. 也就是说，如果您确实经常面临比基线高出 5-10 倍或更高的需求峰值，那么您确实可能是云的良好候选者。这毕竟是 AWS 背后的最初动机。亚马逊需要在黑色星期五或网络星期一比今年其他时间有更多的表现。因此灵活的硬件是有意义的。

But you can also mix and match. The saying goes “buy the baseline, rent the spike”. Many companies won’t even need to bother with that, though. Just buy some powerful machines a fair step ahead of your growth curve, keep an eye on usage over time, and if you have to make an unplanned expansion, you can usually have a whole other fleet of servers online in a week or so. 但您也可以混合搭配。俗话说“买基线，租尖峰”。不过，许多公司甚至不需要为此烦恼。只需购买一些功能强大的机器，比您的增长曲线提前一步，关注一段时间内的使用情况，如果您必须进行计划外的扩展，通常可以在一周左右的时间内在线拥有一整套服务器。

Workfeed goes to America — Workfeed 前往美国

What I found interesting about employee scheduling software is how localized much of it is. You need integrations with payroll and other employee systems to make inroads, and many of those solutions are anchored in the particulars of local labor laws, customs, and industry standards. It's quite unlike the universality afforded Jason and I working on something like Basecamp. 我发现员工排班软件的有趣之处在于它的本地化程度。您需要与薪资和其他员工系统集成才能取得进展，其中许多解决方案都基于当地劳动法、海关和行业标准的细节。这与 Jason 和我在 Basecamp 等项目上工作所提供的普遍性完全不同。

That also means entering a new territory is much harder. Workfeed got started in Denmark, quickly became the top-rated system there, blazed past a bunch of stale competitors, and built a profitable, successful business primarily on the basis of this market alone. But Denmark is tiny. Less than 6 million people. So since day one, I've been pushing them to get ready for the biggest software market in the world: America. 这也意味着进入新领域要困难得多。 Workfeed 在丹麦起步，很快成为那里最受好评的系统，超越了一批陈旧的竞争对手，并主要在这个市场的基础上建立了一个盈利的、成功的业务。但丹麦很小。不到600万人。因此，从第一天起，我就一直敦促他们为世界上最大的软件市场：美国做好准备。

Negative visualization in practice — 实践中的负面形象化

The point of visualizing the consequences of an accident, the loss of a loved one, or the bankruptcy of your business is not to stew in despair, but to prepare yourself for the inevitability of hardship. Nobody makes it through life entirely on easy mode, and if you haven't done any manner of mental preparation, even small setbacks can seem calamitous. So you really ought to prepare. 想象事故、失去亲人或企业破产的后果的目的不是让自己陷入绝望，而是为不可避免的困难做好准备。没有人能完全以轻松的方式度过一生，如果你没有做好任何心理准备，即使是很小的挫折也可能看起来是灾难性的。所以你真的应该做好准备。

In this way, negative visualization is like the dojo in The Matrix. It's a way to level up your skills inside a simulation running within your mind's eye. A chance to realize that "there is no spoon" in the sense that a given external event forces a given internal response. With enough training, you can disconnect the two, and become far better at choosing your response to almost any stimuli. In my book, that's even better than knowing Kung Fu! 这样，负面想象就像《黑客帝国》中的道场。这是一种在你的脑海中运行的模拟中提升你的技能的方法。有机会认识到“没有勺子”，即给定的外部事件会迫使给定的内部反应。经过足够的训练，你可以将两者分开，并且能够更好地选择对几乎任何刺激的反应。在我的书中，这甚至比了解功夫更好！

In addition to building up your resilience, negative visualization also offers a path to gratitude. However bad things might seem in the moment, you can usually imagine something even worse, which should make the present seem far more tolerable. 除了增强你的适应力之外，消极想象还提供了一条感恩之路。无论当下的事情看起来多么糟糕，你通常都可以想象更糟糕的事情，这应该会让现在看起来更容易忍受。