Cloud Hosting (Longterm deployment strategy)

[sacdallago]

initial list

• Digital Ocean
• Heroku
• A2 Hosting
• Linode
• Azure @sacdallago

Factors to consider

• Cost
• Level of support
• Ongoing maintenance costs
• Level of service (uptime, reliability)

Comment on this issue explaining how the services address the factors above.

[jvwong]

Digital Ocean

Pricing

Basic Droplets

Memory	vCPUs	*Transfer	SSD Disk	$/HR	$/MO
4GB	2vCPUs	4TB	80GB	$0.030	$20
8GB	4vCPUs	5TB	160GB	$0.060	$40

• Droplets include free outbound data transfer, starting at 1,000 GiB/month for the smallest plan....Transfer allowance is pooled cumulatively at the account level, not individually at the Droplet level.

General Purpose Droplets

Memory	vCPUs	*Transfer	SSD Disk	$/HR	$/MO
8GB	2vCPU	4TB	25GB	0.089	$60
16GB	4vCPUs	5TB	50GB	$0.179	$120

Service Level Agreement

DigitalOcean provides a 99.99% uptime SLA for both Droplets and block storage. We refund lost time back to your account at the hourly rate incurred.

Support

24/7 Support
Have a question? Send us a note using the form below and someone from the DigitalOcean team will be in touch soon.

Usability

Sign up now with a $100 credit.

( Probably need to try it out to see what's what. )

---

Linode

Pricing

Shared Plans

RAM	CPU	Storage	Transfer	Network In	Network Out	Price
4 GB	2 Cores	80 GB SSD	4 TB	40 Gbps	4000 Mbps	$20 / mo
8GB	4 Cores	160 GB SSD	5 TB	40 Gbps	5000 Mbps	$40 / mo

Dedicated CPU Plans

RAM	CPU	Storage	Transfer	Network In	Network Out	Price
4 GB	2 Cores	80 GB SSD	4 TB	40 Gbps	4000 Mbps	$30 / mo
8GB	4 Cores	160 GB SSD	5 TB	40 Gbps	5000 Mbps	$60 / mo

Service Level Agreement

Uptime Guarantee. We guarantee 99.9% monthly Covered Services uptime
Subject to the Terms of Service and this SLA, the Account Holder may request a pro-rata credit for any affected monthly billing period in which we are unable to satisfy the Uptime Guarantee (collectively “Service Credits”).

Support

Call Support 855-4-LINODE
Chat on IRC
Contact Linode Sales 844-869-6072
...

Usability

( Probably need to try it out to see what's what. )

---

Heroku

Pricing

The containers used at Heroku are called “dynos.” Dynos are isolated, virtualized Linux containers that are designed to execute code based on a user-specified command.

Dyno Type	Memory (RAM)	CPU Share	Compute	Dedicated	Sleeps
hobby	512 MB	1x	1x-4x	no	no
standard-1x	512 MB	1x	1x-4x	no	no
standard-2x	1024 MB	2x	4x-8x	no	no
performance-m	2.5 GB	100%	12x	yes	no
performance-l	14 GB	100%	50x	yes	no

Dyno Type	Price per dyno/month
free	$0
hobby	$7
standard-1x	$25
standard-2x	$50
performance-m	$250
performance-l	$500

Service Level Agreement

Support

Standard
Hours: 12x5
Select a topic, then pick a category and we'll show you commonly asked questions and answers.
Looking for more help? Ask the community or create a ticket to get it routed to the best person to answer it.

Usability

( Probably need to try it out to see what's what. )

---

A2 Hosting

Pricing

Unmanaged VPS

Plan	Runway 1	Runway 2	Runway 4	Supersonic 8
RAM	1 GB	2GB	4 GB	8 GB
RAID-10 SSD Storage	150 GB SSD	250 GB SSD	450 GB SSD	150 GB NVMe
Transfer	2 TB	3 TB	4 TB	2 TB
Cores	1	2	4	2
Reneal Cost /mo	$11.47	$19.12	$25.20	$76.53

Service Level Agreement

Our 24/7/365 support staff with expert system admins keep your server running at its peak. We use industry best practices, the best data centers and ensure your server has the latest security patches. That's why we can commit to your site being up over 99.9% of the time.

Support

Submit a Support Ticket

Usability

( Probably need to try it out to see what's what. )

[maxkfranz]

Re. usability, it looks like we'd have a setup similar to today. A VM would have to be manually managed using docker compose for deployment.

One alternative is to include the app deployment offered by Digital Ocean, where they deploy a docker per Github repo. That requires one VM per docker image. For Biofactoid, that would multiply the monthly cost (factoid 1x, grounding 2x, convertors 3x, rethinkdb 4x, ...). In this case, each app gets automatically deployed when the master branch is updated on Github.

[jvwong]

I know we had Digital Ocean on our list anyways to try out, this just serves as due diligence.

[maxkfranz]

I tried out DO on Friday. The simplest way to do it there would be to just spin up a single VM for Biofactoid prod. We'd start it up manually using docker compose, as we do on the existing cluster. We could make it easier to maintain the DO VM by setting a @reboot cron job so that the dockers spin up when the VM is rebooted. That way, anyone can just hit the restart button in the DO panel if there's a problem.

[jvwong]

What do you think we need in terms of specs to run our app?

[maxkfranz]

I tried the basic one, and it was pretty slow because of the grounding indexing. It was swapping all the time. We probably need a higher option with more RAM for Biofactoid. I suspect we need at least 8GB, maybe 16GB.

[jvwong]

This indexing thing is the biggest pain. Before the rethinkdb dump/restore scripts, I had this thing that dumped and restored the elasticsearch index. I want to revive that if possible - that way, we could just store the data (zenodo) and pop it in...This would reduce our requirements too.

[maxkfranz]

It would be good to have prebuilt indices in any case, but it might not reduce the requirements too much. For something like the grounding search to be fast, you want to keep all the DB in memory if possible. The data is >10GB, so we may end up needing the 16GB option anyway. I expect the minimum would be 8GB even with prebuilt indices.

[sacdallago]

I have had really good experience with them: https://www.hosteurope.de/en/Server/Virtual-Server (the 16GB/RAM version is 40E a month, and you got root -- it doesn't "scale" as easy as AWS/Azure/..., but you can still upgrade and round robin ;))

Slightly different, but consider reading through https://github.com/zszazi/Deep-learning-in-cloud?s=09#deploy-your-model-as-a-web-app and subsequent sections.

[sacdallago]

Ohter plus w.r.t. other things I've seen around:

• No traffic cap
• decent # of cores (although virtual)
• can create backups

Only problem for you guys: hosting in Germany (aka, some latency for Canada/US)

[jvwong]

Testing observations for Digital Ocean. From sign up to running biofactoid via Docker Compose, took less than 45 minutes, which includes setting up users, ssh keys, installing Docker/Compose. This doesn't include restoring data or grounding-search data, however. Really good docs.

Droplet

8 GB Memory / 160 GB Disk / TOR1 - Ubuntu 18.04 (LTS) x64

Indexing:

Source	Time to update (min)
ChEBI	1
NCBI Gene	1
UniProt	6

[jvwong]

Additional considerations:

• Proxy (nginx)
• Data backups
• Certificates/renewal

[jvwong]

'To the cloud' (Digital Ocean) task list:

• Provision DO Droplet (8 GB Memory / 160 GB Disk / TOR1 - Ubuntu 18.04 (LTS) x64)
  • Droplet owned by team 'biofactoid'
• Run Dockerized instance
• Self-restart
  • proxy
  • webapp
  • index
  • grounding
  • converter
  • db
• Certificates: Certbot/Let's Encrypt
• GitHub actions and the DO CLI for semi-automated deployment
• Database backup
  • Approach? Backup strategy rethinkdb-backup#2
• Point master.factoid.baderlab.org at DO Droplet
• Point biofactoid.org at DO Droplet
• SSH to automate deployment to the droplet

[jvwong]

Docker stats for master.factoid.baderlab.org instance on Digital Ocean:

NAME	CPU %	MEM USAGE / LIMIT	MEM %
index	0.53%	2.739GiB / 7.775GiB	35.23%
grounding	0.03%	156.2MiB / 7.775GiB	1.96%
watch	0.00%	6.832MiB / 7.775GiB	0.09%
proxy	0.00%	5.793MiB / 7.775GiB	0.07%
webapp	1.40%	331MiB / 7.775GiB	4.16%
kibana	0.22%	278.5MiB / 7.775GiB	3.50%
dbbackup	0.02%	137.8MiB / 7.775GiB	1.73%
heartbeat	0.03%	12.96MiB / 7.775GiB	0.16%
converters_container	0.09%	179.9MiB / 7.775GiB	2.26%
sync	0.01%	83.47MiB / 7.775GiB	1.05%
db	0.05%	424.3MiB / 7.775GiB	5.33%

Disk:

Filesystem	Size	Used	Avail	Use%
/dev/vda1	167G	40G	127G	24%

RAM:

	total	used	free	shared	buff/cache	available
Mem:	7961	4606	729	1	2625	2968

Here’s what each column means:
total - This number represents the total amount of memory that can be used by the applications.
used - Used memory. It is calculated as: used = total - free - buffers - cache
free - Free / Unused memory.
shared - This column can be ignored as it has no meaning. It is here only for backward compatibility.
buff/cache - The combined memory used by the kernel buffers and page cache and slabs. This memory can be reclaimed at any time if needed by the applications. If you want buffers and cache to be displayed in two separate columns, use the -w option.
available - An estimate of the amount of memory that is available for starting new applications, without swapping

[jvwong]

Summary of instances:

Name	Provider	Domain(s)	DockerHub tag	Check for updates
development	Donnelly	unstable.factoid.baderlab.org	latest	every 15 min
		grounding.baderlab.org
		biopax.baderlab.org/convert
staging	Digital Ocean	master.factoid.baderlab.org	production	-
production	Digital Ocean (TODO)	biofactoid.org	production	-

[jvwong]

Plan for moving biofactoid.org from 'Donnelly' to 'Digital Ocean'

Currently, on Digital Ocean (DO) we have:

• floating IP 174.138.112.180 intended for Droplet biofactoid-staging
• floating IP 159.203.55.214 tied to a Droplet biofactoid-production

Below, we describe a process to test a full biofactoid install on biofactoid-production and migrate the (www.)biofactoid.org domain to biofactoid-production.

• Install and access biofactoid-production via test domain
  • Create a factoid-docker-config branch (i.e. 'beta') that supports the domain 'beta.biofactoid.org'
  • Create test domain 'beta.biofactoid.org' and point at floating IP (159.203.55.214) linked to biofactoid-production
  • Trigger GitHub install workflow (beta branch) for 'beta.biofactoid.org' on biofactoid-production (DO)
  • Restore the latest RethinkDB dump on biofactoid-production (DO)
    • Backup to Dropbox /backup/beta OK?
    • Restore local DB from backup OK?
• Migrate biofactoid.org to biofactoid-production
  • Serve a **maintenance page for (www.)biofactoid.org (Donnelly)
  • Point (www.)biofactoid.org at floating IP (159.203.55.214) for biofactoid-production
  • Trigger GitHub install workflow (main branch) (DO)
  • Restore the latest RethinkDB dump on biofactoid-production (DO)
• Cleanup
  • Delete the biofactoid-staging Droplet, but keep the floating IP (174.138.112.180) (DO)
  • Point beta.biofactoid.org at floating IP 174.138.112.180, intended for biofactoid-staging, eventually
    • Disable/remove master.factoid.baderlab.org so it doesn't point at 174.138.112.180 (biofactoid-staging)
  • Disable CRON jobs for RethinkDB dump (X.X.X.195) (Donnelly)
  • Update cerbot to remove migrated domains (X.X.X.153) (Donnelly)
  • Update internal DNS for relevant domains (www.)biofactoid.org (Donnelly)

---

** Maintenance page