Many of you may have noticed the large-scale outage at Alibaba Cloud recently. As an engineer who frequently works with servers, I’ve worked with quite a few cloud services over the years. Today I’d like to share my personal experience using these providers after all these years.
Let’s start with Alibaba Cloud. This provider was arguably the earliest in China and also the first cloud service provider I personally encountered. If you’ve watched my earlier videos, you might know that when I first started around 2012–2013, this company still had a large number of physical servers. It wasn’t until later that physical servers could no longer be renewed and we had to purchase their cloud servers instead. Of course, back then, cloud servers weren’t very stable — website outages and network fluctuations were common. However, after so many years of development, things have become quite stable, especially for their earliest cloud server products. Over the years, besides cloud servers, Alibaba Cloud has also developed many other cloud product lines, such as cloud databases, object storage, and various networking and security products. In short, anything you can think of that an enterprise might need is continuously being migrated to the cloud.
Looking back over these years, Alibaba Cloud’s overall failure rate has been acceptable. Why do I say “overall” failure rate? Because Alibaba’s product portfolio is simply too vast — the probability of failures in certain services or nodes does exist. For example, last year when our company was using their cloud database product, we encountered a situation where even the most basic disk was filled up, leading to log errors. However, that kind of issue at most affects a few servers or certain clusters. An epic, hours-long large-scale outage like the one Alibaba experienced a few days ago is very rare — you might not see it once in several years. For Alibaba Cloud, which has been developing for so many years, the impact is enormous. For instance, several companies I know that deeply rely on Alibaba Cloud products were all affected to some degree this time. So much so that if we previously thought the gap between Alibaba Cloud and AWS was gradually narrowing, now we have to seriously reconsider that gap.
Of course, in terms of after-sales service, Alibaba Cloud’s support is still quite good. Although it has been streamlined from the original ticket-plus-phone support to the current model where phone callbacks are only provided if you specifically request them — otherwise you’ll only get ticket replies — fortunately, most issues can be resolved through tickets, community articles, and online resources. This is actually quite understandable: as the user base grows, responding to everyone by phone would be far less efficient. That said, many smaller service providers, in order to carve out a piece of the competitive server market, still offer both phone and ticket support.
Tencent Cloud
A provider with a noticeably different after-sales experience from Alibaba Cloud is Tencent Cloud, which we’ll talk about next. Especially in the earlier years, their product pages were rather confusing, and you could clearly sense that their customer service staff didn’t really understand their own products — it was obvious that you were being served by an outsourced support team. For some questions, the support agent would even need to consult other support staff.
Tencent Cloud appeared relatively later and also experienced significant outages in earlier years, even losing user data on one occasion. Fortunately, in recent years it has developed quite well, and the cloud business has received more attention, gradually becoming more stable and mature. Tencent Cloud’s pricing is generally somewhat cheaper than Alibaba Cloud’s. For example, our company previously used Qiniu Cloud’s object storage, but due to high costs, we later negotiated with Tencent Cloud’s sales team and got a good price, subsequently migrating some data to Tencent Cloud.
Regular users tend to be more price-sensitive — a difference of a few dozen yuan between cloud providers might lead some users to compare back and forth. Their stability requirements aren’t as high either; even occasional minor issues won’t cause significant losses. Many cloud service providers also offer bigger discounts targeting new individual users.
For enterprise users, however, it’s a different story. Every minute of business downtime means not just lost revenue but also reputational damage. Besides placing special emphasis on stability, pricing is even more important for enterprises than for individual users. Individual users purchase relatively small quantities, but enterprise users, as their business grows, need more and more cloud products across more categories — beyond just cloud servers, there are cloud databases, cloud security products, log analysis, and so on. Costs naturally need to be kept within a reasonable range. Of course, purchasing in larger quantities also gives you more bargaining power. For enterprise users, more often than not, costs can be reduced through agencies or direct negotiation. For example, in companies I’ve previously worked at, we basically secured discounts through bulk purchasing, through agencies, or even by becoming an agent ourselves. The prices were much cheaper than buying directly from the official website — many products could get discounts of 30% or even more, and there would also be annual rebates for the enterprise.
The final point I want to make is this: regardless of which provider you choose — whether public cloud, private cloud, or a hybrid of both — you should never place everything with a single cloud service provider. This is to ensure data security and service stability. Nor should you become overly dependent on any single provider.
As the saying goes, it’s easy to migrate to the cloud but hard to leave it. Many enterprises initially find cloud services extremely convenient, and certifications in many areas are also easier — for example, for the Ministry of Public Security’s Information Security Protection (Děngbǎo) certification, you only need to submit screenshots of purchased products and usage records as supporting materials. You only need to handle maintenance at the business level. However, once an enterprise reaches a certain scale, they realize that cloud service costs are simply too high, and their business has become deeply dependent on cloud services. At that point, breaking away from cloud services is really difficult — it requires modifying too many things and will also sacrifice some stability. Speaking of which, when we see news like Elon Musk’s Twitter moving off the cloud, it becomes much easier to understand. After moving off the cloud, it’s reported that compute costs alone were reduced by 50%, and overall costs dropped by 60%.
Video references:
https://youtu.be/QJz8Oyz-chw
https://www.ixigua.com/7309764024986501659?logTag=93c2dfdb917c43e0ae5a