Category Archives: tech

macOS Allow CLI binaries from third-party devs

01 Friday Apr 2022

Posted by in aide-memoire, tech

≈ Leave a comment

Tags

macOS Big Sur (apparently from macOS Catalina onwards) restricts binaries from third-parties. This sux when using cli apps like alfred-plugins.

Allow CLI app: spctl --add /full/path/to/my/app

Show all rules: spctl --list

Disable Gatekeeper: sudo spctl --master-disable

Another was to achive the same result: Hold ⌥ (option) when opening app in Finder; this will fail but give you the option to “open anyway”. If you forget to hold ⌥ or it does not work, try opening anyways, then open System Preferences -> Security & Privacy -> General; click the Open Anyway button.

Rating: Bad Apple.
Good intentions (methinks), but Apple’s gonna lock down macOS, just like it locks down iOS today. 😔. This is one more arrow in their quiver.

Advertisement

Enterprise / Server Cloud Storage (not “Personal” cloud storage)

02 Saturday Jan 2021

Posted by in aide-memoire, programming, tech

≈ Leave a comment

Cheapest:

  • Backblaze B2: $5/Tb/mo (data transfer: $10/Tb)
  • Wasabi Storage: $5/Tb/mo (data transfer: ??)
  • OVH physical server: $3/Tb/mo (zero data transfer and other charges, but you do need to setup ZFS and some API layer software yourself).

Backblaze and Wasabi are both S3 compatible object stores. Note that data transfer can easily dominate storage charges, for even somewhat frequently accessed data. OVH is therefore much cheaper than it appears.

Other Options

  • google / gcloud: $10/Tb/mo
    • 1 Tb  => $10/mo ($120/pa)
    • Nearline (USA): 1000 * 0.01 => $10/mo (Using prices for nearline storage, South Carolina data center)
    • Nearline (Mumbai): 1000 * $0.016 = $16/mo ($200/pa)
    • Nearline has some charges if stored for less than 30d (min term). So could be a bit more than $10 (perhaps the same as S3’s intelligent tiering)
  • amazon / s3
    • 1 Tb => $12.5/mo  (between $12.5 and $20 because intelligent tiering will take some time to kick in).
    • Using prices for S3 Intelligent Tiering, USA data center: 1000 * 0.0125 => $12.5
  • Digital Ocean / volumes: (Block storage built on Ceph)
    • 1 Tb => $100/mo (approx 10x S3 prices, yiiikes.)
    • rsync’able storage
  • Digital Ocean / spaces: $0.02 gb/mo (S3 Compatible File Storage)
    • 1 Tb => $20/mo (about the same as S3 without intelligent tiering)
  • Backblaze B2: $5/Tb/mo
    • $10/Tb data transfer; cheaper than Amazon, but can still dominate.
    • S3 compatible API

Max value

Store HOT on Digital Ocean spaces (save on bandwidth)

Move older items to gcloud nearline or lower tier

#[[Hosting / Servers / Cloud Guide]]

“Free” SQL Databases

29 Tuesday Dec 2020

Posted by in aide-memoire, programming, tech

≈ Leave a comment

  • Heroku Postgres
    • Good & simple
  • IBM DB2: No credit card required, re-activate by email every 90 days
    • 200 MB of data storage
    • 15 simultaneous connections
    • Shared multitenant system
  • “Always free” Oracle Cloud instance (ptooie, slow as molasses) + apt install postgres
  • Supabase: Postgres, realtime, kinda like Firebase but with SQL and a real postgres instance. Free while in beta
  • Recommendations:

(Low) code API integrations

19 Wednesday Feb 2020

Posted by in finance, programming, tech

≈ Leave a comment

Tags

What does a distributed / internet operating system look like?

A platform that lets you natively read/write from modern apps? Make apps?

In some sense this is what IFTTT was going for – though consumer focussed, it is IMHO, a very example of a ‘no-code‘ development platform . It’s what Apple’s ‘Shortcuts’ aka Workflow is. It’s what Alexa skills are. I’m not sure Airtable fits, but Airtable got the marketing right.

And now we’re seeing the green shoots of a new generation; pipedream.com and autocode.stdlib.com. These are more developer focussed and provide the “plumbing”, but still need devs to actually make stuff. I’m a lot more excited (perhaps that’s cuz I’m a dev, and these platforms are targeted right at me). I think there’s going to be an explosion of new apps + services, because it’s just so, so easy and quick to knock out an app that uses say Twitter and Amazon S3 … is this how the promise of “serverless” will finally be delivered?

+QQQ

(Virtual) Server Pricing

04 Tuesday Jun 2019

Posted by in aide-memoire, tech, Uncategorized

≈ Leave a comment

My goto services:

#FREE => Heroku (but only one “VM” is free, then $7/mo for a 512Mb “Dyno”)

Super Cheap => OVH
don’t have an affiliate program AFAIK, cuz they’re so low cost 😦

Blend => Digital Ocean
Use my aff. link to sign up, you get $100, I get $10, everyone’s happy 🙂
Here => https://m.do.co/c/728c8434e1b1

Digital Ocean
Special sauce: Incredible UI/UX, super fun to use. I’ve stopped thinking about it anymore, when I need a VM, I just go to DO.

Downsides: Many (probably most?) of the IPs are blacklisted for email. Most aren’t great exits for VPN services either for the same reason.

CLI tool (doctl) has pretty shit UX; could have been so much better. But still it’s there and works in a pinch		$5/GB of RAM
Linode$5/GB of RAM
OVH
Special sauce: The lowest cost option, but it’s Canada and Europe only. No SE Asia, no USA.		$1.68/GB of RAM
Yep it’s 66% cheaper than everyone else.
Vultr
Special sauce: awesome plans with $2.50 for a node with IPv6 only, 512Mb RAM. 		$5/Gb
Amazon LightSail
Special Sauce: oh well, it’s the IBM of servers, so I might as well document it here: They have this clone of Digital Ocean, called LightSail, that s3cks, obvs. It’s the same price as Linode/DO/Vultr. Available in a few more locations than most of the above (incl. Mumbai; Digital Ocean has Bangalore).

When to use: All the rest of your infra is on Amazon Cloud and you’re pathethic.		$5/GB of RAM
Heroku #FREE (owned by Salesforce, not indie and chill like they pretend)
Incredibly good tooling; 1x 512Mb Dyno is free (1000 Hours/mo if you verify with a credit card). Gets very expensive very fast though — because you are subsidising the development of developer tools and the “free” tier.		$14-$50/Gb of RAM
Prices as of June 5, 2019.

I may update this page someday, but then again I may not 🙂 Bookmark if you’re adventurous (ctrl + d, or cmd + d)

Online IDEs for quick prototypes

09 Tuesday Apr 2019

Posted by in aide-memoire, lifehack, programming, tech

≈ Leave a comment

Tags

#Frontend

Codepen.io: Really great for prototyping layouts and CSS in general.
Recently added Flutter support

  • React: ✔
  • Svelte: x
  • Flutter: ✔
  • JS: ✔ (but not very good …)
  • Private Pens: ✔ (Paid)
  • Secrets: x
  • Deploy to Netlify (or now.sh): x

Codesandbox.io: Webpack + VS Code as an online IDE

#Fullstack (include a VM/Container for development)

Glitch.com: #fullstack, #millenial, #awesome, #secrets

Glitch + Fly.io is very fly!

Repl.it: #fullstack, #secrets, #paid-for-private-repls
Has secrets (.env file that is only accessible by repl owner)

Runkit: The best way to hack nodejs ‘scripts’, IMHO
#backend

Codeanywhere, $8/mo

Stackblitz:
VSCode in the Cloud

#secrets  = .env file that is only accessible by repl owner

Archived

C9.io: Was really nice, but it’s been bought by Amazon; not free anymore, unless you use AWS 😭

Do you know others that should be on this list? Like, subscribe and Leave a comment below!

 

Docker ❤️ Parallels ❤️ OS X (El-Capitan)

11 Wednesday Nov 2015

Posted by in aide-memoire, programming, tech

≈ Leave a comment

Tags

, , ,

Getting started with Docker + Parallels on OS X (using Homebrew)

Update: Sep 2016: This is no longer relevant; best to download Docker for Mac which uses xhyve to virtualize the Docker Engine environment and Linux kernel-specific features on OS X. This allows for much easier sharing of files between your local machine and docker containers.

Want to use Docker on OS X? At the moment, the OS X kernel does not directly support containers like Docker or OpenVZ. So you have to run a VM with docker installed. You then use the Docker CLI to interact with the Docker (daemon) running on the VM. The whole process is actually very easy to setup now that Docker Machine supports the Parallels driver.

Assumptions: You already have Parallels (11, Business or Pro Edition) and Homebrew installed.

$> brew update  # need to brew with a recent version
$> brew upgrade
$> brew install docker docker-machine docker-machine-parallels
… grab some coffee…
$> docker-machine create -d parallels dev1
$> eval `docker-machine env dev1`
$> docker run hello-world

All done!

You are now ready to use Docker, Docker Composer and even Docker Swarm.

Apple shipped 3x as much battery capacity as Tesla in 2014

25 Wednesday Feb 2015

Posted by in finance, tech, trading

≈ Leave a comment

Tags

, , ,

So, everyone’s talking about the possibility of an Apple iCar; it looks certain that it’s got something to do with batteries (Apple’s being sued for recruiting scientists from A123) and arch-rival (wannabe arch-rival?) Samsung just bought out a company that makes automobile batteries…

So I was thinking – who sells more batteries today? Apple or Tesla? And because it’s not quite Apples to Apples I’ll try comparing battery-capacity.

Apple Battery Capacity 2014:

  • approx. 180m iPhones; the iPhone 6 has a 7.9 watt-hour battery, while the 6+ has 11 watt-hours. Presumably a decent chunk of the 180m phones were older 5, 5s, 4 etc. So let’s say the average capacity per iPhone is 8 watt-hours. Total: 180m * 8 = 1440 million watt-hours (i.e. 1.44 GWh aka Giga-Watt-hours)
  • approx: 60m iPads; the iPad Air 2 has a 27.6 Wh battery, while the Air (1) had a 32.9 Wh battery. Let’s hand-wave and say 28 Wh average. Total: 60m * 28 = 1680 million watt-hours (1.68 GWh)
  • approx: 19m MacBooks; MacBook Air’s have between 38 & 54 Wh, Pro’s 63.5 to 91 Wh. Let’s say 54 Wh on avg. Total: 19m * 54 Wh = (1.02 GWh)

Apple also sold a quite a few other devices with batteries, remotes, iPods, Beats headphones, Wireless keyboards and watt-have-you (haha), but’s let’s ignore it for now.

 

So Apple shipped 4 GWh (1.44+1.68+1.02=4.14 GWh) of battery capacity last year alone. That’s about $2 billion (cost) assuming Apple has the same cost as Tesla for battery capacity. I’d argue that it’s lower, but who knows.

 

“Tesla Motors may have the lowest rates for electric car batteries; the estimated battery costs for Tesla Motors is around US$200 dollars per kWh.” — From <http://en.wikipedia.org/wiki/Tesla_Motors>

 

Now Tesla shipped 17,300 Model S cars (http://insideevs.com/monthly-plug-in-sales-scorecard/). They can have a 65kWh or 85kWh battery-pack. Say 75kWh on average. Total: 17.3k * 75kWH = 1.3GWh.

 

So there you have it. Apple shipped 3x as much battery capacity as Tesla in 2014.

 

Some more fun napkin math (let me know if it’s all wrong…): Apple’s batteries pack 25% more energy per gram than Tesla’ (so for the same weight, using Apple battery tech in Tesla’s could improve range by 25%!)

 

The iPhone 5s battery is 26g -> 7Wh 26g (26/7 = 3.7 g/Wh)

(http://www.ebay.com/itm/Original-1560mAh-3-8V-Li-ion-Internal-Replacement-Battery-for-iPhone-5C-5S-/201227137579?pt=LH_DefaultDomain_2&hash=item2eda12722b)

 

Tesla cells are : 10Wh 45g -> (45/10 = 4.5 g/Wh)

(http://industrial.panasonic.com/lecs/www-data/pdf2/ACI4000/ACI4000CE17.pdf)

 

Apple’s batteries probably charge faster too (assuming that you can parallel charge the 1200 or so MacBook Air batteries that make up a single Tesla sized 65kWh pack).

 

Also interesting: Batteries like memory are super-high-margin. Like 60-70% gross.

 

Disclosure: I’m long Apple (+ve net delta) and have no position in Tesla.

Does an $intc short make sense?

30 Wednesday Apr 2014

Posted by in finance, tech, trading, trends

≈ Leave a comment

Job postings suggest Amazon is planning to make its own ARM-based servers.

Google eyes Power chips

Software Engineer, ARM Server ($FB)

Looks like everyone is evaluating ARM servers; makes sense to me. Small, cheap servers that can saturate a 100Mbps network connection…

I think $INTC is under attack on all fronts:

Servers: ARM could make serious inroads (classic Clayton Christensen, the innovation looks like a toy, flies under the price umbrella till it’s too late for the incumbent, who cannot (or will not) self-cannibalize based on something that looks like a toy …)

Mobile: $QCOM, $AAPL, SMSN:LI rule here…

Desktop/Notebook: As I posit in my last article, $AAPL may make a move here (or not). But the fundamental thesis is the same. Desktops will move to from the Core variants to the Atom variants. They’re nearly good enough now. $MSFT providing Windows free for low-end boxen? It’ll only add to the fire.

If the move from Core/Xeon series is towards Atom – it will margin compression. If it’s towards ARM, it will mean substantial loss of market share. And $INTC is now setup to be a player at scale. Not sure if they can compete like a scrappy little startup anymore. I wonder if that’s left in the DNA?

As a long term $INTC fan – nearly every compute device I’ve bought had Intel Inside, this marks the end of an era to me….

When will the transition be visible in stock prices? Is much of this already priced in?

With $INTC trading near the top of it’s 52W range ($27.24) I think the risk (for a short) is quite low. Say about $4  = 15% ($31), which is where many analysts are calling it. The gain could easily be $7 = 26% ($20). I think it’s time to nibble a bit…

Red:

Short interest seems to be under 1% of the float, why is it so low?

PE ratio is just 14; multiple expansion is easily possible if $INTC shows signs of life in the mobile space. $INTC has been talking about a better radio, which seems to be a key weakness for them….

There could be a revival in the desktop space (unlikely IMHO).

The main risk I see is that the move could take longer than I can hold … will a move take more than a year?

Amber:

I just love the Xeon Phi (in abstract terms). Can’t see the whole point of Tesla/GPU compute if you can do cheap enough, dense enough full-featured cores. A whole bunch of ‘em. But then there’s the fun projects like Parallella. The software effort required for to write apps for GPU compute languages has never made sense to me.

Green:

NUC is a joke.

Atom in the Datacenter means severe margin compression. ‘Course an Atom doesn’t measure up to the raw compute of the Core / Xeon E series … but then the price differential is substantial.

What’s the $INTC bull case?

1) Success in mobile – scaling down the Ivy-Bridge architecture to be competitive in the 2W TDP space.

I think $INTC is competitive already – but it’s not leading to design wins because:

1.1) Low-end is extremely price-conscious; INTC realistically has not much   chance of competing in this space. Sure they can ‘bribe’ companies (contra-revenue as they’re calling it) to use INTC, but I think this is akin to buying eyeballs….pointless in the medium term.

1.2) High-end, there’s no market outside of AAPL, SMSN and neither is going to use INTC at the moment. Perhaps Nokia (err… Microsoft Mobile Oy) may give it a shot? Lenovo? maybe, but really I can’t see a third player (at any scale) for at least the next few years.

Can INTC use the leverage it has – in terms of desktop/server to bully Apple? Don’t think so — Apple is perhaps one the largest single customers for INTC, they need Apple more than Apple need INTC.

2) Datacenter growth continues un-abated, with Intel continuing to rule the roost, picking up another couple of percent in terms of market-share by destroying what little remains of AMD?

I certainly think Datacenter will continue to grow pretty fast. The cloud is more important than ever. Indeed INTC bet that growth in mobile would require equivalent growth in datacenter and the margins in datacenter were much fatter….(find quote from outgoing INTC CEO here…). Is the relationship linear?

Cloud Storage

19 Wednesday May 2010

Posted by in lifehack, tech, trends

≈ Leave a comment

[EDIT: Update]

Wrote this post thinking about Google Docs. Turns out Google is copying Amazon and has come out with its own “Google Storage for Developers”. Basically identical to Amazon, just that it costs $0.17 per GB per month instead of the $.20 per GB per month that Amazon charges. Data transfer in and out costs a heck of a lot of money, especially to-from APAC ($0.30 / GB)!

Which means running your own storage server (in a well built and well managed data-center) is probably the best bet. A decent server with decent storage say, 2Terabytes of disk (1TB x2) costs about $100 per month which is … $0.1 GB/month with RAID level 1 – Mirroring or $0.5 per GB/month with RAID level 0 – striping. In all practicality there will be no further data-transfer charges.

‘Course you have run your own linux distro, but you can choose whatever super cool filesystem you want. If you use Kimsufi.co.uk you can run Proxmox VE (or VMWare or Hyper-V) + a pre-packaged appliance like Openfiler.

[EDIT: Original Post continues]

Google is nearly an order of magnitude cheaper than Amazon S3 for cloud storage!

  • Google: Max 1GB, Crappy interface for large files, tools ecosystem not as rich as S3.

  • Amazon S3: Max 5Gb per file. Decent tools. Have to pay for good usermode filesystems. EBS allows you to easily store/mount true-crypt encrypted volumes.
    • 80 Gb for $192 p.a (full redundancy)
    • 80 Gb for $96 p.a (reduced redundancy storage)

And Google has no data-transfer charges … which can be very, very substantial! Amazon is free for inbound data till June 2k10 …

I think I’m going to start transitioning my content from S3 to Google. Going to try and store my everyday desktop as a VM image – let’s see how that goes.

BTW: Noticed that Reliance has substantially lower latency than Tata when it comes to international bandwidth. Round-trip times under a 100ms for servers in Paris/London. Which means remote desktop may actually be usable. I remember a senior (and brilliant, though understated) colleague, saying that anything under 200ms is ok for UI tasks … (this is when I was developing embedded devices and we were trying to work out how fast we needed the processor to be …). Methinks Tata goes via SGP => TKY => NYK to reach LDN (they acquired Tyco). Reliance on the other hand owns FLAG and can go via the Middle East. When I’m next on a Tata line, I’ll check out ping times to NYK. Should be about the same either way to NYK (~250ms), but LDN should be 400ms via Tata if I’m right.

I think Amazon has servers in SGP. I think Google is more of HKG play – which will add some 50ms to round-trip times.