| Region | Type | Notes |
|---|---|---|
ap-south-1 |
ra3.xlplus |
The unusually slow disk-read-write benchmark has returned to normal. |
ap-southeast-3 |
dc2.large |
Disk-read benchmark is very slow, at 0.49s, rather than the usual ~0.06s. |
ca-central-1 |
dc2.large |
The disk-read benchmark has after a month returned to normal (0.06s), from being about 0.7s. However, disk-read-write and networking are both now slow, although still in the normal range. |
eu-central-1 |
dc2.large |
Networking has a high standard deviation, at 1.02s, part of the slowest networking benchmark seen for this region, at 5.73s. Usually it’s about 4.0s to 4.5s. |
eu-south-1 |
dc2.large |
Having a slow day for disk-read, at 0.18s (rather than 0.06s). |
eu-west-2 |
dc2.large |
Also slow on disk-read, 0.17s. |
us-east-1 |
dc2.large |
Slow on disk-read, 0.13s. Also slow last benchmark, at 0.15s. |
us-west-1 |
dc2.large |
Slow on disk-read, 0.17s. |
https://www.redshiftresearchproject.org/cross_region_benchmarks/index.html
Just spent an hour trying to resize the disk on a KVM/QEMU VM.
Tried and failed.
The phrase “impenetrable mess” is duelling with the phrase “incoherent mess”, in my mind.
Disks, partitions, LVM, physical, logical…
qemu-img, fsblk, pvs, lvs, pvresize, lvresize, growpart, fdisk…
I could resize the KVM disk image just fine.
Getting beyond that to resizing the LVM physical volume, and then the LVM logical volume?
Googling gives a buzzing, blaring confusion of information and misinformation.
An hour later, I’m no further forward, and I’ve given up.
It’s going to be quicker and easier to make a new VM and install Debian from scratch.
(I don’t need to configure Debian - that would be a lot of work - just get some software installed that I’ll use once).
I’ve been using computers all my life, and I’m a serious programmer. If I can’t get LVM to work, it’s probably too complicated for its own good.
I wanted to write a bit about the supply of cluster munitions to Ukraine.
I may be wrong, but this seems absolutely the right and correct step to take.
There are four reasons.
Russia invaded Ukraine, and is subjecting the people in occupied territories to State-organized death, torture and oppression. The military effort by Ukraine is the only way to get those people out of that situation; the downside of unexploded cluster munitions must be weighed against the goal they are being used to achieve.
The Ukrainian offensive is, as offensives do, consuming vast amounts of artillery ammunition, about 100,000 rounds per month - and that’s with the Ukrainians rationing use. There is an intense need for more artillery ammunition. The problem is that Ukraine has used up its stocks (indeed, Russia for years was sabotaging ammunition dumps in Ukraine, and by this Ukraine lost far more ammunition than it actually had at the start of the invasion), NATO has limited stocks and has provided much of them already, and production takes at least a year, and really two or three, to ramp up.
There are three million cluster-type artillery rounds in storage in NATO.
It’s a case of using it, because supplies of conventional ammunition are unavailable.
The Russians have been using cluster munitions since day one of the invasion, over 500 days ago - and their ammunition has a much higher failure rate than Western ammunition. (The failure rate is an issue in that unexploded bomblets remain in the ground, and sometimes manage to work in the end, killing people years later.)
The fighting now is focused on a narrow band, maybe 5 to 10 kilometers deep, of Russian fortifications, which the Ukrainians are trying to break through, so they can use manoeuvre warfare. Unexploded bomblets will be focused on this narrow regions, not across the country as a whole.
ra3.xlplus is showing a halving
of performance over a number of regions.| Region | Node Type | Notes |
|---|---|---|
af-south-1 |
ra3.xlplus |
Disk-read benchmark running at about half speed. |
ap-northeast-3 |
all |
This region looks to have been upgraded. I
benchmark dc2.large and ra3.xlplus and
performance has moved from the “slow region” group into the “fast
region” group. However, this happened once before, in 2023-05, but
lasted only for that one benchmark. Additionally, despite the upgrade,
the disk-read benchmark for ra3.xlplus can also be seen as
running at about half speed. |
ap-south-1 |
dc2.large |
Disk-read benchmark running slow, at 0.27s, rather than the usual 0.06s. |
ap-southeast-3 |
dc2.large |
Disk-read benchmark has returned to normal (from 0.49s). |
eu-north-1 |
ra3.xlplus |
Disk-read benchmark running at about half speed. |
eu-south-1 |
ra3.xlplus |
Disk-read benchmark running at about half speed. |
eu-west-2 |
dc2.large |
Disk-read benchmark running slow, at 0.68s, rather than the usual 0.06s. Was also showing slowness last two previous benchmarks. |
me-south-1 |
ra3.xlplus |
Major problems; disk-read running
half-speed, like many regions, but disk-read-write and networking
benchmarks are both very slow - about a quarter of normal
performance. The dc2.large type is normal. |
us-west-1 |
dc2.large |
Disk-read benchmark running slow, at 0.34s, rather than the usual 0.06s. Was also showing slowness in the previous benchmark. |
Interesting question on Stack Overflow;
Why do I get different results with regexp_replace when run inside a proc vs outside of a proc in Redshift?
The problem, I think, is that Redshift does not use the text of queries as is - as you enter it.
Considerable modification occurs, and there are a number of bugs in those modifications, and the modifications vary inside and outside of procedures, and it is for example the modified text which is stored into the system tables (which log query texts).
It looks to me like the author is experiencing text modification which is removing spaces which ought not to be removed.
The solution is going to be a work-around, whereby you store the text into a table, and use that text in your query by reading it from the table.
I’ve gradually increasingly been thinking I ought to write a bit more about the Ukrainian offensive; I do not follow the news, as it is really entertainment, but I am keeping an eye on the Russo-Ukraine war, and so I’m seeing some of what’s being written in the press, and by and large, what’s being written, to my eye, is uninformed and by that confused.
When it comes to offensives, what you want to do is get mobile forces into open territory so you can make large enveloping movements, cutting off and encircling large bodies of enemy forces.
To begin with though it can be that you’re facing a solid defensive line, and you need to get through that line, to get mobile forces into open territory.
That line is going to be ten, twenty, even thirty kilometers thick.
Within in there’s a main defensive line, where the defences have been made as strong as possible, and ahead of that, for tens of kilometers, are lesser defences, to slow and grind down the enemy.
In all of this you must remember the enormous power of the defence. It is much easier to defend, than to attack.
So to begin with, on offence, you need to grind your way through the outer defences, just to get to the main line of defence, and then you need to get through that, and on a wide enough front you can support the mobile forces you intend to push through into open territory - many tens of thousands of troops need a lot of supply, which means plenty of roads.
Back in WW2, after the German collapse in France, the Allies rushed eastwards, but eventually lost momentum, due to supply constraints, and due to the recovery of the German defence.
Once the front had stabilized, the Allies found themselves in much the same position as the Ukrainians now; a solid defensive line, which they needed to grind through, to get mobile forces into open territory.
It was taking ages. Weeks, months - you grind, the enemy reinforces, you grind more, gradually disrupting the defence, you push in multiple places to divide the defence, it all takes time.
As I understand it, the Ukrainians have arranged their forces into three groups. The first group are to push forward until the main defence line, the second group are to get through the main defence line, the third is the mobile group which will deploy in open territory.
All of this is absolutely correct tactically.
What we’re seeing now is simply that defence is powerful and it takes a long time to grind through tens of kilometers of minefields covered by fire.
Home 3D Друк Blog Bring-Up Times Consultancy Cross-Region Benchmarks Email Forums IRC Mailing Lists Reddit Redshift Price Tracker Redshift Version Tracker Redshift Workbench System Table Tracker The Known Universe Twitter White Papers