WikiFur:Moving hosts
From WikiFur, the furry encyclopedia.
WikiFur is moving hosts in the near future. This page explains why, and explores the various options available to us.
Contents
Summary
Since its creation in July 2005, WikiFur has been hosted on Wikia, a full-service wiki hosting provider, or wiki farm.
Wikia is not a charity. It is a for-profit company funded by venture capital. This has many positive aspects - not least the high performance and availability of WikiFur to date. However, over time, the negative effects have become more evident.
On Friday 6 June 2008, Wikia announced that it would be enforcing a new default skin (Monaco) with significantly increased advertising, some of which would intrude into the content area (in the top-right position normally occupied by a picture or infobox). Users would still be able use other skins by logging in and selecting them in their preferences, but administrators could not select a non-Monaco default for users. They have also discarded a promise made before WikiFur's arrival to never host popup ads (this change was reverted after further discussion).
We believe that this new style and level of advertising negatively impacts our readers to an unacceptable level. We also feel that Wikia is asserting an inappropriate level of ownership over the wikis for which it provides services. They have indicated that they want a more cohesive community of wiki sites than we do. These differences of opinion and needs are not likely to be resolved.
WikiFur is not the only site to have problems with this - it has been the subject of much discussion. Members of Wookieepedia (the Star Wars Wiki) have come to the conclusion that they need a "plan B". We concur, and intend to use it as soon as possible.
Administrative support for above statement
- Support
- I'm GreenReaper, and I approve this message. --GreenReaper(talk) 20:07, 14 June 2008 (UTC)
- Agreed. Moving not only frees us from the reigns of Wikia, but opens up numerous new opportunities for WikiFur as a whole. --— beeps (talk, contribs) 21:10, 14 June 2008 (UTC)
- Agreed. I think that a move would help us grow and is in our long-term best interests. --Douglas Muth 21:53, 14 June 2008 (UTC)
- With Wikia's intent now clear, both in their desire to increase monetization by advertising alone rather than by sales of services, and in their desire to claim greater ownership over member wikis, I feel that WikiFur will be best served by moving to an independent host. -- Siege(talk) 15:10, 15 June 2008 (UTC)
- I also am strongly in support. -- Sine 01:31, 16 June 2008 (UTC)
- I agree. Establishing independence from this sort of money-grubbing is definitely in WikiFur's long-term interest, especially if Wikia's changes lead to attempts at editorial control - a definite no-no for a wiki. Carl Fox 03:13, 16 June 2008 (UTC)
- I support a move. I dislike pop-ups. I looked at the wikia's communitytest of the Monaco and while I think I could handle one of the Google ad boxes, a double-dose of Google ads at the top of every page looks like it would be ad-heavy. --EarthFurst 04:01, 16 June 2008 (UTC)
- Unfortunately, it looks like Wikia won't compromise in its decision so it's necessary to find a new host. --Rat 07:30, 16 June 2008 (UTC)
- I'm going out to get boxes and parcel tape, pop-ups are just unacceptable and moving to a new server is quite feasible. --Nidonocu - talk
13:22, 16 June 2008 (UTC)
- Not giving a choice to the customer (us) means not caring about them. Also, pop-ups make me cringe! Either point is enough for me to want to move. • Ekevu • ♥ • ★ • 14:16, 16 June 2008 (UTC)
- I agree Spirou 02:21, 17 June 2008 (UTC)
- I concur with the reasons stated for finding a new host. It's been good with Wikia so far, but I don't like where they are heading. --mwalimu 21:30, 17 June 2008 (UTC)
- I support the move - finding a new host does seem to be in the best interests of WikiFur's future.--Higgs Raccoon 20:54, 18 June 2008 (UTC)
- Neutral
- Oppose
- Recuse
Things to consider
- Wikia may still maintain a site at furry.wikia.com. This is up to them - the GFDL permits it, although certain images uploaded under different licenses (or none) might not be available for their use.
- For this reason, please immediately change links to use en.wikifur.com rather than furry.wikia.com. The new domain will work both before and after the move.
- GreenReaper has control over the WikiFur service mark, therefore our site will be using it and Wikia will not.
- A full database backup (not just the XML) and image backup should be available to us. User login details may be available.
Current status
Our images currently occupy 1.4GiB of space. The database page table can be compressed if necessary and in that case should take up less than 100MB. Allowing for reasonable expansion over the next year, we will be at 2GiB.
Our total bandwidth use is hard to calculate, but we use about 50GiB a month in CSS and HTML transfer alone. It is not unrealistic to suggest that our total usage might be three times this after images.
Comparison of options
During the administrative discussion, several options were proposed.
Some options that were quickly dismissed:
- Wikia has indicated that while they might consider an offer of money for hosting, it does not really fit into their business model. WikiFur's administrators felt it would not work on a long-term basis.
- Other wiki farms were ignored, as we do not believe they will offer a sufficient level of service to justify the additional cost.
NearlyFreeSpeech.net is a shared hosting provider with load-balancing. They provide Linux, PHP 5 and MySQL. They host Anthrocon's website (and GreenReaper's personal site).
- Pros
- We only pay for what we use - no more, no less
- Integrated rsync.net option for a low price with no transfer costs
- Should provide good availability and probably sufficient performance
- Competent techs willing to "go the extra mile" for the right people (as certified by Giza)
- Cons
- The costs provide a disincentive to grow our audience and our site:
- The cost of bandwidth, while great for small sites, becomes uncompetitive quite quickly, even with the long-term bandwith discount
- Storage costs are high - $10.00 per GiB/month. No bulk discounts (yet).
- Limited control over the environment. We can run CGI scripts manually (remote triggered), and there is ssh and [S]FTP, but we can't add or configure OS-level modules ourselves.
- No ability to run most others kind of servers (no static IP).
- No memcached (in-memory object caching daemon), which could severely limit MediaWiki performance
- Estimated cost
- Storage: 2.0 * $10.00 * 12 = $240
- Bandwidth: 150 GiB * 12 = 1800 GiB, which will cost $479.82. The subsequent year would be $407.12.
- MySQL: ($0.01 [base process] + $0.01 [InnoDB table support]) * 365 = $7.30
- Backup: 2.0 * $1.40 * 12 = $33.60 (optional)
- Total: $760.72 for the first year, with backup (varies significantly depending on accurate estimate of bandwidth usage)
Virtual server (Linode.com)
Linode offers Linux virtual servers occupying high-powered machines (quad-core Intel Xeons with RAID-1 disk mirroring), with a certain dedicated amount of memory, storage and bandwidth, and a guaranteed (small) percentage of the real computer's CPU which may increase if other sites are not using their share.
- Pros
- Even the smallest plan should have sufficient storage for us
- Location would likely be in New Jersey, a good compromise for American and worldwide furs.
- Real IP address, real server inasmuch as we can run what we like on it, full control.
- Cons
- MediaWiki can be rather CPU-expensive as an application, and may also strain the memory. We are already close to the limits of the lowest plan. We can buy more RAM and bandwidth, but that will also cost more.
- We would have to provide our own server administrator and spend more time running the server - time we could be spending improving or promoting the wiki.
- Estimated cost
- 360 plan (max sharing: 40, 360MiB RAM, 12/18GiB storage, 200 GiB/month bandwidth: $19.95 * 12 = $239.40 a year
- 540 plan (max sharing: 30, 540MiB RAM, 18/27GiB storage, 300 GiB/month bandwidth: $29.95 * 12 = $359.40 a year
Use of dedicated server (Timduru)
Timduru (our French WikiFur administrator) has offered the use of his dedicated US server (Conroe 3040 dual-core, 2GB RAM, 750GB+500GB HD, 3TB/month, ~1.8TB free - FreeBSD 7.0) hosted at ThePlanet in Texas. This is the same one that hosts a number of fursuit-related sites (e.g. The Fursuit Archive, The Fursuit Database). It is also a streaming server for the Funday PawPet Show and FursuitTV.
- Pros
- Low-latency, relatively high-bandwidth (40MBps+) connectivity with a monthly limit that we are unlikely to max out anytime soon.
- Server facilities appear relatively unused despite the existing services. There's plenty of CPU and memory available, even during periods of streaming. Initial tests indicate performance would be very good.
- Ability to run daemons as required by the site, and perhaps other servers. SSH access. Our own IP.
- A server administrator who intimately understands both our needs and the server.
- A backup to his French server can be provided, and potentially activated there in an emergency.
- The price is unbeatable.
- Cons
- Timduru has been a server administrator for 10 years, but the people he has to rely on to fix problems have sometimes been less than fully competent. While the control situation is now improved (a remote KVM is installed) we would still have to act on any low-level problems via Timduru.
- There is the potential for degraded network performance during periods of streaming. This does not actually appear to be the case but it is possible that this will occur as viewership continues to increase.
- Timduru's server is primarily for supporting and promoting fursuiting within the fandom. If push came to shove, WikiFur would be the site to be shoved - gently, with a reasonable time to arrange alternative hosting. (This situation is considered "very unlikely", as increased supply of bandwidth/storage/etc. has outpaced demand in recent years.)
- Estimated cost
- Nothing, unless we required services with a specific cost. However, as a significant user, we might well chip in now and then, and/or direct user donations to help cover yearly costs, which are currently in the $2000-3000 range.
Vote
- This vote will close at 6AM Thursday UTC (1AM EST, 10PM Wednesday PST)
Please indicate your preference by adding your vote to this table. If you have no preference between two options, average your vote. For example, if you preferred one option above two others, vote 1 for the first, and 2.5 for the others.
- This vote is of an advisory nature. If there is a clear preference for an option, it will be taken. If not, GreenReaper will make the final choice.
- The vote is open to all users, but the votes of administrators may be given more weight.
| Option 1 (NFSN) | Option 2 (Linode) | Option 3 (Timduru) | Signature/comments |
|---|---|---|---|
| 2 | 2 | 2 | No comment. --Example 20:07, 14 June 2008 (UTC) |
| 2 | 3 | 1 | I believe that Timduru's solution is the most effective in our case. Linode - as documented - will likely have problems running the MediaWiki software and would require an administrator to be available in the area in case something does go awry. NFS is not only expensive, but limited in it's application and control we have over the operating environment. Meanwhile, Timduru's option has the least severe "cons", while we can be assured that the interests of Timduru and WikiFur are aligned and understandable. And hey, better the devil you know. — beeps (talk, contribs) 21:11, 14 June 2008 (UTC) |
| 1 | 3 | 2 | I feel that NFSN is a good First Step, as it is easy to get started and they have a robust environment. If they fail to meet our needs, then Timduru would be the next best, due to cost (free) and the fact that there is already a server admin. --Douglas Muth 21:57, 14 June 2008 (UTC) |
| 1.5 | 3 | 1.5 | While I feel that Timduru may be our best option, I also feel that Mr. Muth has made a good case for NFSN. Linode is good for monetary cost, but needing our own server admin is an incidental cost we might not be able to afford right away. -- Siege(talk) 15:36, 15 June 2008 (UTC) |
| 2 | 3 | 1 | I'm quite sceptical about vhost providers. They often don't scale well. Mediawiki without shell access can be quite a PITA for certain admin tasks, thus I'm not conviced that NFSN suits us. Timduru, though, is a very reliable and experienced system administrator. o'wolf 15:58, 15 June 2008 (UTC)
|
| 3 | 1.5 | 1.5 | Timduru: NFSN infrastructure sounds interesting, and based on other sites hosted there it sounds like a good option, it is quite costly though. The main points that worries me about them is that as written in their FAQ, they don't allow processes to run for more than 2 to 5 minutes for shell and 3 minutes for php. Mediawiki maintenance scripts and import / export processes will take more time than that to complete, and we will need to run them on a regular basis. The initial xml import took more than 2.5 hours to complete on my server. Even if we get a DB export, that'll take more than 2 minutes to complete I think. The other aspect written in their FAQ is that it seems they don't expect you to use a lot of ressource as you would do on a dedicated server, otherwise they might cancel your hosting. If these 2 points could be cleared with them so that we are sure we can run a heavy mediawiki there, I think their redundant servers would be better than mine. About Linode it's difficult to know what power we would get, but based on the server specs it might be enough. The good point would be that we have full access to the server too. --Timduru 17:53, 15 June 2008 (UTC) |
| 2 | 3 | 1 | Unless something really cool pops up re: NFSN, I would be inclined to go with the resource we're more familiar with. :) Carl Fox 03:23, 16 June 2008 (UTC) |
| 1.5 | 1.5 | 3 | I haven't got a detailed understanding of the technical requirements, so I won't try to figure out what would be better between NearlyFreeSpeech and Linode. I am uncomfortable with the scenario of being hosted as a favour. Sine 04:55, 16 June 2008 (UTC) |
| 3 | 2 | 1 | We're going to need a server admin no matter which we pick, so we may as well pick one familiar with the server. Rat 07:30, 16 June 2008 (UTC) |
| 2 | 3 | 1 | As per Rat's comments. Also, I agree with o'wolf as in virtual servers don't scale well; they're meant for starting up, which isn't our case. • Ekevu • ♥ • ★ • 14:13, 16 June 2008 (UTC) |
| 2 | 1 | 3 | I don't really like any of the options presented, but the option for a dedicated server is the one I dislike the least. The first option isn't bad either, but discouraging the growth of the site is probably a bad idea. And, with all due respect to Timduru, hosting something like Wikifur on someone's personal server, with a single admin, is a very bad idea. This is why I didn't offer my hosting services, and I even have active staff besides myself, with more system resources than Timduru's quoted specs. Felix Softpaw 23:42, 16 June 2008 (UTC) |
| 2 | 2 | 2 | Abstain. I will leave the final decision about this matter to more knowledgeable Admins. I can't bring anything to this discussion at present time Spirou 02:21, 17 June 2008 (UTC) |
| 2 | 2 | 2 | Abstain. I've read and pondered the pros and cons for each choice, but I just don't have any experience in hosting and maintaining a large website, and I think the decision is best left to those with expert knowledge.--Higgs Raccoon 20:54, 18 June 2008 (UTC) |
| 3 | 2 | 1 | While I don't really know a whole lot about servers, the third option has the most advantages in it's favor. And the only thing that I worry about is server stress (streaming) as we continue to grow. This just seems to be the best option available right now. We do of course have the other two as backup options. --Markus(talk) 21:38, 18 June 2008 (UTC) |
| 3 | 2 | 1 | I agree with Markus, in fact I would write something very similar. -- Xkun(talk) 22:33, 18 June 2008 (UTC) |
| 3 | 1 | 2 | Well, I would essentially rule out NFSN due to the nature of their hosting platform. Reasons for this include: Very limited server access, no possibility to host other services in the future, no persistent processes (which means we'd have to run mediawiki as a pure CGI program, which is very very very slow), and no memcached or squid (which is absolutely needed for WikiFur with it's level of activity). This leaves Timduru's offer and Linode.
|