|
关于运输问题:- I can confirm that Tokyo will have multiple servers sent via UPS 1-3 day shipping instead of the original 7 or so day air freight. We'll send enough to meet the entire immediate demand, so this will catch us back up to the original timeline. We're going to consider doing the entire batch with priority shipping and not freight, unless we can get UPS freight to be responsive again (or FedEx.)
复制代码 我可以确认,东京将通过UPS 1-3天的运输发送多个服务器,而不是原来的7天左右的空运。我们将发送足够的服务器来满足整个即时需求,所以这将使我们赶上原来的时间表。我们将考虑用优先运输而不是货运来做整个批次,除非我们能让UPS货运再次有反应(或联邦快递。)
关于网络:- I'll just post all Tokyo subnets here since it will at least provide a full picture if people are that interested and unofficially spreading this information. 176.119.148.0/22 78.142.228.0/22 88.214.20.0/22 213.232.112.0/22 147.78.240.0/21 45.66.128.0/21 Keep in mind none of these are set up on the network equipment yet because it's not there yet, so they will naturally fail ping or otherwise have 100% packet loss. We can't fulfill any specific subnet requests. As for the networking itself, I have a private test IP on my end that the datacenter provided from another cabinet in the same datacenter with the same network blend, and I've been testing it against a lot of random IP addresses in China on China Telecom, Unicom, Alibaba, Tencent, etc, and in various regions. The general performance is what's usually expected of Tokyo, it will be low ping, but there will still be some packet loss. I actually misplaced all the data I recorded, let me see if it's on my other computer, but it was around 58ms ping to Shenzhen Tencent I believe (I may be misremembering) and very low packet loss. China Telecom had probably 12% packet loss to Shanghai or ... actually let me just try to find it, my memory's not that good, I don't want to throw out random numbers (but I mean this is the act of throwing random numbers, this isn't scientific by any means.) Japan will be able to burst up to 10Gbps as a whole (not per server, not all servers are 10Gbps.) The average usage per server is around 40Mbps in general, across all our regions all the time (if we average everything over months, including backups.) The backups will no longer be done as often over the network so that number will go down unless we double bandwidth for everyone (just kidding, double bandwidth would basically literally have zero effect, it's just a marketing gimmick.) Anyway, the current commit is above 40Mbps per server so we do not expect any kind of major congestion on our end, at least no more than current levels, even with a lot of users.
复制代码 我就把所有的东京子网贴在这里,因为如果人们有那么大的兴趣和非官方地传播这些信息,它至少可以提供一个完整的画面。
176.119.148.0/22
78.142.228.0/22
88.214.20.0/22
213.232.112.0/22
147.78.240.0/21
45.66.128.0/21
请记住,这些都还没有设置在网络设备上,因为它还没有出现,所以它们自然会ping失败或有100%的数据包损失。我们不能满足任何具体的子网要求。至于网络本身,我有一个私有的测试IP,是数据中心从同一数据中心的另一个机柜提供的,具有相同的网络混合,我一直在与中国电信、联通、阿里巴巴、腾讯等公司的很多随机IP地址进行测试,而且是在不同的地区。
一般的性能是通常预期的东京,会有低的ping,但还是会有一些丢包。我实际上把我记录的所有数据都放错了地方,让我看看是否在我的另一台电脑上,但我相信到深圳腾讯的ping大约是58ms(我可能记错了),丢包率非常低。中国电信到上海大概有12%的丢包率,或者......其实让我试着找一下,我的记忆力不是很好,我不想乱扔数字(但我的意思是这是乱扔数字的行为,这决不是科学。)
日本将能够整体爆发到10Gbps(不是每台服务器,不是所有的服务器都是10Gbps。)每台服务器的平均使用量在一般情况下是40Mbps左右,在我们所有的地区都是如此(如果我们在几个月内平均所有的东西,包括备份。)备份将不再通过网络经常进行,所以这个数字将下降,除非我们把每个人的带宽加倍(开玩笑,加倍带宽基本上没有什么影响,这只是一个营销噱头。) 不管怎么说,目前每台服务器的投入都在40Mbps以上,所以我们并不指望在我们这边出现任何形式的重大拥堵,至少不会超过目前的水平,即使有很多用户。 |
|