3Kir Kolyshkin wrote on the OpenVZ blog, Why Virtuozzo is good for OpenVZ,
The first version of Virtuozzo was released about 6 years ago, so it is not something new. Virtuozzo costs money, and is used by big corporations for mission-critical applications. Virtuozzo customers expect it to be very stable, fast, bug-free, well documented, etc. And to sell the product successfully, those expectations must be met.
Thus he concluded,
Without Virtuozzo, OpenVZ quality would definitely be lower. A dedicated quality assurance team is always needed because it’s just not appealing for developers to find a new bugs in their code — they’d rather write some more.
Okay. Let us take his description of Virtuozzo, and substitute it with Linux.
The first version of Linux was released about 16 years ago, so it is not something new. Linux costs money, and is used by big corporations for mission-critical applications. Linux customers expect it to be very stable, fast, bug-free, well documented, etc. And to sell the product successfully, those expectations must be met.
Hmm. That sounds right to me. You can in fact swap the term “Linux” with other products like “Apache“, “Samba“, etc (you do have to adjust the years though), and the paragraph will still make sense.
Oh wait! Maybe not. Linux, and its list of GNU tools actually do not cost money — you can freely download the source, compile them, modify them, redistribute them, etc. The same can be said about all other major open source projects that Kolyshkin himself adores. Suddenly, the necessity of having a fast and stable proprietary software to back up the open source version no longer holds.
D’oh.
I still appreciates SWSoft opening up part of its source code and supports OpenVZ. I am also a direct beneficiary as we have been using OpenVZ at work to set up developers’ boxes, so a team of 8 can happily share a single Dell PowerEdge 2950.
However at the same time I can see quite a few short comings in OpenVZ that have already been addressed by Virtuozzo, but are not likely to come to OpenVZ anytime. Most of them are UBC related, where I see Virtuozzo’s SLM memory model a fit solution.
In a forum discussion on WHT, I commented:
OpenVZ claims to be the virtualization and automation basis of Virtuozzo, but I wonder whether SLM will ever be rolled into OpenVZ. It certainly makes OpenVZ feeling like a “free but limited” product of Virtuozzo, than “cutting edge and experimental” proving ground of Virtuozzo that they are trying to make people to believe.
Matt Ayres from Tektonic/Unixshell replied:
If OpenVZ included VZFS and SLM I know I’d have written a control panel for it and ditched Virtuozzo by now :) That is very bad for SWSoft. We must appreciate that they are a business that wants to make a profit and have been generous by providing such a complete open source virtualization platform in the form of OpenVZ.
Well said.
Is Virtuozzo good for OpenVZ? If you are a open source/free software purist, then No!! And please go and hack support platforms like Linux-VServer to make it more suitable for hosting.
Otherwise I’ll argue that Virtuozzo is a necessary evil for OpenVZ, from business point of view.
Interesting to mention Linux VServer. I know the project very well since it’s beginnings and even contributed lots of testing and some code to it’s earlier versions. I am actually amazed OpenVZ was adopted so rapidly by many while Linux VServer/FreeVPS projects have both existed for years before. Linux VServer is a more advanced kernel than OpenVZ’s in some areas with CoW links vs overlay filesystem (VZFS), a limit for RSS (even before Virtuozzo SLM was available), and CFQ disk I/O QoS implemented per-VPS. What OpenVZ did provide was mature tools, easy installation, and a 100+ page PDF manual. It doesn’t help that Linux VServer does not yet have iptables inside the VPS or even 2nd-level quota support (which it actually had in it’s 2.4 kernel). Now based on all of this, why did you choose OpenVZ over VServer?
Hi Matt!
“Now based on all of this, why did you choose OpenVZ over VServer?”
Sold by the SWSoft propaganda that Virtuozzo == VPS?
We actually have used OpenVZ for quite a while at work when we started getting those 4-core 2850/2950’s with massive amount of RAM for development. When I did a search on “VPS” on Google, it looked like the most obvious choice was Virtuozzo because “everyone else was using it”. As OpenVZ is similar to Virtuozzo so we went ahead with it.
I was aware of VServer but had an impression that quite a lot of things have not been virtualized (networking, or iptables as you have mentioned), but from the spec it does have a lot of advantages over OpenVZ. It’s definitely on my to-do list, but I’ll leave it to next year :)
Matt, aren’t you taking simlpy what Herbert tells you?
1. Have you looked at so called disk I/O QoS in Vserver?
It is not per-VPS. Simply run 10 tars in one VPS and
monitor how it affects another one. You will be suprised :)
You will be suprised to know that CFQ scheduler also doesn’t handle writes.
OpenVZ is going to fix all this very soon and the first step was already done -
commited disk I/O accounting patches.
2. limit for RSS? do you really think that killing an application w/o giving it a chance
to handle out-of-memory situation gracefully is a good idea?
ok, if I wanted to select the easies way, I would definetely choose the same.
why don’t you use vserver if it is more advanced? don’t you want to do all this dirty hacks and tweaks to make it work?