Everything Apple (the Mac heads thread)

I have a feeling it may be passive as I don't think that 24 inch needs 20+ Gbps data, but it does need to provide power. Could be wrong though.

The first gen 5K is TB3 only but the new gen have USB C as fallback.

I may be wrong about the first bit, it is kinda confusing.

I looked at iCloud, but really only wanted one cloud solution as an ongoing cost which ended up being Dropbox. The price of the 1TB external SSD drive works out about 2 years of iCloud, and I should get many more years out of it.

Has anyone set up a nas as seperate drives rather than a raid?
Put two 8tb drives into a nas and want to use them to back up my media nas and also use as a time machine for my Apple desktop.
Would prefer to have each on a seperate drive so I could swap out individually if they fail.

Cables longer than 0.5m have to be active AFAICT.

The 4K display supports TB3 pass through intended for a second display over a single cable from the computer. If it didn't it could happily do everything over a passive USB-C cable.

Nah, there are loads of passive 2m TB3 cables... I'm pretty sure... yeah.. https://www.amazon.co.uk/Certified-Cable-Matters-Thunderbolt-Supporting-Black/dp/B01AS8U9KE/ref=asc_df_B01AS8U9KE/?tag=googshopuk-21&linkCode=df0&hvadid=310842873031&hvpos=&hvnetw=g&hvrand=14739633716356716529&hvpone=&hvptwo=&hvqmt=&hvdev=c&hvdvcmdl=&hvlocint=&hvlocphy=9044936&hvtargid=pla-441004577250&psc=1

I have an iPhone, iPad Pro and Mac Book Pro, and have 2 Terabytes of iCloud storage, ticks the “just works” box. It automatically tides stuff off the various devices to free up local storage, and means that documents are available across devices. It also ensures that they’re all backed up all the time, so can be restored easily if needing replacement or significant servicing.

I bought the additional iCloud space to share with my family. Well worth the cost and hassle free.

This is (presumably) off-site replication rather than backup. It's not advised as a single backup method as if the hardware fails your backups are as screwed as your primary drives. You can do it if you want a hot standby, but you're trading off reliability for availability as NAS 2 is always on. If you go down this route, make sure you use the right combination of rsync options to ensure that you don't automatically delete files from your destination server in the case that a file is deleted from the source server. You can run a warm standby but it's not a substitute for backing up your data.

i have JBOD, single drives and RAID 10 in my NAS but I still back up to external HDs, not ones mounted and managed from within a NAS OS.

That was the original intention but the enclosures the drives came in didn’t power on. Brought the nas assuming I could replicate that.

I've run replication across 2 NAS's but as I say, it's for redundancy of data availability rather than a bombproof backup.

Imagine the following scenarios:

1. house is broken in to and computer equipment gets stolen
   
2. NAS1 will not power on - fully dead hardware.
   
3. You accidentally delete an important file and do not notice for 6 months.
   
   

Which of these would replication to a second NAS solve?

Additionally, NAS2 is equally likely to fail, giving you a false sense of security about your data integrity.

If you think the internet is a passing fad or have very sensitive data then backup to cold storage, in a fireproof safe, ideally offsite. Otherwise backup to a reliable cloud subscription service. There are lots

Just out of interest and because you might know. How do the cloud services ensure they don't suffer data loss?

They make backups.

What salad said.

Except moar.

They do have backups, and backups of backups (because the backups are also on their storage so get backed up). And they have replication even without backups, so live data exists in multiple places on multiple servers in multiple datacentres.

There are abstractions on abstractions that help answer a simple question like: Where is this file, can I get it please? And this gets mightily complex when you introduce caching and all of the places a file may additionally exist in addition to the replicated storage, backups and backups of backups.

To say that there are at minimum 3 copies of a file is to underestimate how many times a file exists, and just how much of an outage can really make the file unavailable.

Storage and systems are no longer the worry for such companies, it's networking and the "ooops, a minor outage has caused a domino effect of outages because we accidentally sent all of the network traffic to some small rack of servers".

The problems may on the surface look similar between a home user and a cloud company... but different parts of the problem are solved for the cloud company and the problem is very different. There's virtually nothing a cloud company is doing that you could apply at home, i.e. how many multiple servers and networks do you have? Are they even in different physical locations?

Oh, and the other real problem for cloud companies is SSD failure rate.

It's high.

SSDs do not last long at all.

And whilst there is a lot of spinning discs, most of the hot layers (cache, heavy hitter files) are on SSDs and SSDs fail at a spectacular rate.

You may think your SSD is super reliable, and better than a spinning disc. But SSDs are rated for number write operations, and in a cloud environment they are kept very busy with writes and their life is barely measured in years even though they all exceed their rated number of operations. In a big enough fleet, that means SSDs are being replaced daily, in very high numbers.

Granted I don't have a clue just how they actually break - is it possible to recycle them?

It's not really possible to recycle them. They all get shipped eventually to parts of India and China where low paid labour just pulls chips off of the circuit boards for classification, and these are later sold at a fraction of the cost.

By and large though... no significant recycling happens to any of the tech that the human race produces, it all goes to landfill. The cost of recycling being greater than the cost of the item to produce again. All of the efficiency that exists, exists in the production of these things at a low cost with acceptable pollution and externalised costs... there's virtually no efficiency in resource re-use and preservation, recycling, etc. If you want a capitalist society with a liquid economy and to be able to afford most of what you consume and purchase, this is what that takes today.

As to how they break: Their memory can only take a number of writes before it fails... enough bits failing and the SSD can no longer find places to put files. So they degrade whilst bits are failing, and once enough bits have failed they degrade faster as fewer bits now receive all of the writes. As soon as the OS can no longer write to an SSD the drive is pulled and discarded and a new one added. If multiple SSDs in a server died (replication exists within a single server too) then the server itself is removed from the fleet (networking stops sending traffic to it) and all of the SSDs are replaced (if the server is a high cost one) or the server is just permanently routed around until the vast majority of its' rack has died - and then the whole rack is replaced at once (a cheaper operation if your fleet is cheap commodity hardware and you have a lot of it).

^ fascinating. Thanks for the write up!

As to how they break: Their memory can only take a number of writes before it fails... enough bits failing and the SSD can no longer find places to put files. So they degrade whilst bits are failing

This seems to suggest the total capacity of the drive would "shrink" the more it's being used.
I always thought they retain full capacity ..but then just go die at some point and that's that.

They retain their full readable capacity... but in most places they are used the data changes.

i.e. within a cache layer the algorithm applied might be LRU (Least Recently Used), which means to put new items into the cache the oldest items are evicted. Eviction here just means over-written... but if you can't write, then for the purpose of cache your storage is reduced as you can't add new items to the cache.

Example on LFGSS... the attachments. The attachments live in an LRU cache, and new attachments in hot (most visited) conversations will be requested the most. So those items are added to cache, and because cache is always full (it's expensive SSDs and less capacity than long term storage) the least recently served (oldest) item in cache is deleted from cache to make space for the new item. If I can never delete the oldest, then eventually I can never add newest, making the cache server useless to me (because requests would be cache misses and reach the far slower spinning disks somewhere... I care about speed, as slower reads take longer, resulting in contention, which floods networks - and networks are the problems cloud providers really have. Fast caches, SSDs, save networks.

But if your SSD is full of photos that you never delete... you're good, they're still there for you as read-only.

Thanks for elaborating, appreciate it! 👍

Any suggestions for somewhere in London that does short term rental of Macpro or iMac Pro. Searching around a lot of companies look quite dodgy.

Dry or wet? Short term as in for a day/couple of days for a shoot or for a few weeks?

I can recommend some DITs that come with kit but not sure about dry hire only.

Dry hire, just need a machine for me to use. Short as in days or maybe 2 weeks.

Somewhat cheeky, but 2 weeks is well within Apple's returns period...