Last Week on My Mac: The arithmetic of backing up

After Macs themselves, storage and backup hardware accounts for the largest part of my hardware budget. It’s also the gear that I plan most carefully: my current Time Machine backups go onto a wonderful Promise Pegasus R4 RAID system which is coming up for replacement. So I’m trying to plan its successor from scratch.

The first question is how big that backup storage needs to be, something that is fundamental to any decision and one of the hardest numbers to predict. On the face of it, all it should need is sufficient capacity to make its first and full backup, in my case around 1.5 TB, plus sufficient space for at least a couple of years of added backups, say another 1 TB. That though depends on the longevity of the storage: it’s probably wisest to replace hard drives after about three years when their warranty expires, and no more than five, but SSDs should be good for nearer ten.

There’s a side-effect to periodic media replacement though. What I’ve done in the past is pop the hard drives from my RAID and put them aside as archives. That in turn relies on replacement not being so costly that I’ll want to delay it as long as possible, which depends on whether I’m using hard drives (at around $/€/£ 80 per TB) or SSDs (upwards of $/€/£ 150 per TB).

Either way, total usable storage needs to be more than 2 TB, and if it’s to last longer, it may need to be over 4 TB. So the longer it lasts before replacement, the greater the capacity required, the more expensive it’s going to be, and the more likely I’m going to want it to last longer. That looks like a vicious circle from here.

The other crucial question is whether my future backups can be made to SSDs using APFS, or will be stuck with HFS+ as at present. Whilst APFS does run on hard disks, it sucks badly when away from SSDs, as anyone who has tried booting a Mac from an internal hard disk knows only too well. Neither am I enamoured with the prospect of using HFS+ on SSDs, although in that case, it’s not performance that concerns me, but longevity, as HFS+ was designed for hard drives with their essentially unlimited number of writes, which are key to determining SSD life.

There’s no shortage of external hard drives, including many different models which could support RAID in hardware or software. Branded 8 TB Thunderbolt 3 systems range from around $/€/£ 600 to the successor to my existing Promise, the Pegasus3 at $/€/£ 1400 inclusive of four 4 TB disks.

Looking at SSDs, the options are far more limited. I could get a single Samsung 860 EVO 4 TB SSD for just over $/€/£ 500 plus the cost of a case, but as it’s still a SATA device its performance won’t be that much better than my existing RAID system. To get the best out of Thunderbolt 3 I’d either need to RAID four 1 TB SATA SSDs together, or go to much more expensive ‘blades’.

Enclosures for SSDs are also much harder to come by. For four SATA units there are a few available for $/€/£ 500 or slightly more, including the OWC ThunderBay. Look for anything comparable to the Promise Pegasus3 for, say, four 1 TB SSD blades and I can find just one model, the OWC ThunderBlade, which currently costs $/€/£ 1279 – almost the same as 4 x 4 TB in a Promise Pegasus3.

But the biggest uncertainty in all this is what’s happening to Time Machine. Over the last three versions of macOS, it has been edging closer to version 2.0, using APFS snapshots where it can, and deciding the best strategy for determining what to back up. One reader here has even been able to get the latest version to back up to a sparsebundle copied from network storage to a local APFS-format disk. But that’s still a far cry from being able to make regular Time Machine backups to the new format, which is what we need to make backing up to SSD to make sense.

Even that is only half the story: let’s say that over the next year, Apple were to add support for APFS on local backup destinations. What would happen to existing backups? Would there be a reliable and seamless migration route? Would we still be able to back up to hard disks in HFS+, or would all backup destinations have to be APFS volumes? Lack of any insight into these issues makes any decision a gamble.

So at present, my choices are:

  • replace the hard disks in my current Promise RAID system for less than $/€/£ 300;
  • replace my Promise RAID system with another hard disk RAID system for $/€/£ 600-1400;
  • replace my Promise RAID system with single 4 TB SSD at over $/€/£ 500;
  • replace my Promise RAID system with a 4 x 1 TB SSD SATA enclosure at just over $/€/£ 1100, or a ThunderBlade 4 x 1 TB system at $/€/£ 1279.

The snag is that it may take more than a year to discover in retrospect which was the right choice.