Show newer

Friend: it was so cold other day.

Me: so?

Friend: my laptop froze.

Me: 😦

Friend: it was my fault. I left too many windows open.

Do you need a fun game for the kids for Christmas this year? 😰🤣

This is technically a backup. While cygwin has some performance issue with pipe, I can avoid that by letting dd read the whole disk and write to tape.

It's hard to read (have to dump to a spare disk then pick the file I need), but it still cound as a backup, I guess.

Tried a bunch of method to speed up/buffer the write speed to tape drive in Cygwin. Then found out the pipe itself has performance issues.

With `dd if=/path/to/test/file of=/dev/nst0 bs=4M`, it can easily get 160MB/s writing speed. But with `dd if=/path/to/file bs=4M | dd of=/dev/nst0 bs=4M status=progress`, I can only get about 65MB/s.

That's not usable.

Wrote a hacky bash script using LTOEnc (a windows cli tool for setting up LTO hardware encryption) and tar to backup my PC. (Using Cygwin)

But I'm confused with `tar -b`. With larger blocking factor, each record get bigger but when dealing small files there are more space wasted. I'm backing up both C and D drive, where D drive lives my code and movies/animes/shows.

Don't want to be too extrem on parameters, so I use `-b 1024` and enabled compression to make balance.

Also enabled encryption. While I don't want to encrypt long time archive, backup is a more short term thing, which I'm pretty sure I won't forget the password.

Should be fine I guess? Now I'm testing it.

Want to test how good the hardware compress is. So I decide to test is using block number. I assume the block number will be smaller if compression reduce the total amount of data write to tape.

But then I suddenly realized, that compression is transparent to the system. No matter compress or not, the block number is same since I use the same dataset to test.

So now I'm measuring rewind time😋

hmmm

looks like stenc cannot be compiled under cygwin.

Still, not found any way to enable hardware encryption of my LTO tape drive.

@C99CXMOE 可是我现在仍然买不起雷电3拓展坞(小声

印象中两年还是一年多以前,在fedi刷到一条嘟大意是:加拿大社工以及政府的服务逻辑并不是为了让人们获得资源,而是不让人们得到这些。我当时觉得非常反直觉,隐约觉得能懂但也没太想明白。但工作时间久了接触到的case多了之后完全理解了那段话。我最近才知道这边的工人赔偿(worker's compensation),在claim和拿钱阶段都不能工作或者是有非常严格的限制工作条例,背后逻辑是所谓“保护工人”,既然你受伤了为什么还要去工作?工作了不就说明你可以工作不需要补贴?

但实际上,claim的过程通常非常繁琐非常漫长,在这段时间里你拿不到钱,但是也不能工作,就等于完全断了收入来源。而且不仅不能工作,也不能去上学。我有个case是client因为长期在Child and Family services工作,如果有相关经验的人看到CFS这个缩写都应该知道这个领域压力有多大。她工作的领域非常前线非常高风险,直接面对了caseload中太多死去的儿童之后觉得自己不能继续做了,在supervisor建议下申请了精神工伤补贴。听起来很健全吧?精神工伤补贴都有哦。

但是政府一是刁难认证过程,说你要怎么证明这份工作确实给你带来了精神伤害?于是她花了二是用大量的paperwork和漫长的等待时间折磨人,申请过程非常繁琐,而且精神工伤的claim是很容易失败的,那么等待期的零收入就全是沉没成本。但她还是通过种种努力很不容易地成功了,但是呢,在拿到这笔钱中间又被拖了很久而且仍然有很多钱还没拿到,这期间也还是不能上学工作。期间她申请了一个在线项目,但因为不能上学一再推迟最后取消了。此外,申请工伤赔偿还需要用掉也就是清空all the holiday hours,如果在之后找到的工作里她需要喘息都没有机会再用这个假。在这些漫长的反复拉锯里,我的client说她最明显的感受是“政府是在想尽办法证明我不能得到这些”。可能有些人会觉得,这些都是那种陷入贫困陷入绝境没怎么受过教育的人吧,否则也不会对这些程序都handle不了——我这个client有三个degree。And that's how the system works.

Wait, what?

Cygwin can detect my tape drive and put it to /dev/st0?

Now I'm going to read the 9th chapter of tar manual.

Also I noticed the JVM's impl of SHA3-256 doesn't perform well. It's kind of 150MB/s and can't saturate my tape drive which can read 160MB/s.

I have no issue with Bouncy Castle's impl which is purely in Java.

I thought native impl is better. Why? (I tried Java 21 LTS, no improvement)

“你是否蔑视你的政府?”
“与我的政府对我的蔑视相比 我的蔑视不值一提”

赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢赢

Finished two tape cartridges. One more to go.

But at least I think my software is working and usable. Although I notice HPE include utility tools that functionally similar to mine.

Anyway, for anyone who use LTFS with LTO 5 and later tapes, you may have a try:

github.com/hurui200320/Offline

最近在做跨境汇款的一点点经验总结。
1.尽量不要使用中国银行,如需使用,请提前致电客服查询自己的跨境转账限额,以及国内转账限额。
2.每年五万美金限额是指购买限额,是央行直接统计,按身份证号码统计。你的卖出并不扣减。比如你买了五万卖了一万,再想买一万是不可能的。换银行也不行。
3.如果遭遇转账限额,app提示超出限额,打客服咨询解决办法,如果本人不在国内,客服表示无法解决时,要求对方寻找可解决的主管。并且登记你的问题,给出问题记录号码,会在三个工作日后,有你的开户行工作人员联系你。
4.如果遭遇转账限额,本人在国内,可持本人身份证到网点或开户行解决该问题(提升限额,或临时提升限额)
5.限额问题四大银行均有,目前国内银行大概有钱荒,未来可能会更严重。直接后果是,自己的钱在账户,但是,转出或取出存在障碍。

最后,人如果在国内,我没啥建议,找个信得过的非四大银行?我也不知道。
人在海外就是,钱转出来就ok了。

Despite TeraCopy has a better GUI, it actually faster than ltfscopy.

The ltfscopy simply calls other program like copy (on Windows, or cp on Linux), then call md5sum twice to get original file hash and target file hash, all in sequence. Which I think you can calculate the hash when you copying it, or just calculate the two md5 checksum at the same time.

With a 330GB file, my SSD can read up to 300MB/s, and LTO6 tape drive can read up to 160MB/s. I can save ~18 minutes just by doing it smarter.

PS: TeraCopy do support SHA3-256 but it can only handle ~80MB/s, which is a huge shame. I have to use Blake3 to speed things up.

Show older
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.