Re: [CODE4LIB] Trends with virtualization
I have seen the pendulum swing back and forth several times over the last 20 years between dumb terminals and complete PC's with their own set of apps each. Philosophically, the tension is between control and anarchy; cost is just brought in to justify your position. If you love control, then dumb terminals are what you want. Since this means things are centralized, it requires important hardware and backup systems to make sure it never goes down. I think of this as the nuclear aircraft carrier mentality - sinking a nuclear carrier would be such a catastrophe (to both sides) that you need umpteen other ships to protect it from ever happening. I am more of an anarchist: I have faith in people's innate ability to muddle through okay for themselves. It doesn't bother me so much that people make mistakes and do dumb things; I try to set things up to blunt that, but other people's mistakes really not my responsibility. I try to set things up more on the side of boppo the clown - the weighted blow-up figure that you can keep hitting forever and still have it come back without effort. So I love being able to snapshot VMs before doing anything new; no longer are you risking rebuilding the whole machine every time you update/install something new. VMs let me give people the leeway to shoot themselves in the foot without hurting others. This is a great confidence-builder for people; they will come up with new ways of doing things far more often when the penalties for mistakes are not so severe. The second thing I love about vms is that you can delete them. This is because you can afford to use them for just one or two things. In the old days (pre-2006) when everything was on bare metal, you bought a big machine (aircraft carrier) and put all the business processes on it until there were too many to ever have the server go down. In practical terms, security was non-existent, because no one could ever keep up with which task needed to do what after a while, and no one wanted to screw up some important process that everyone had forgotten needed rights to some files somewhere obscure. So the longer a server lasted, the more extra rights were left over from previous business processes that no one even quite remembered any more. But a VM you can delete when the main business process on it stops. You will have had some security creep unless you really named your groups well, but that all goes away when you kill the VM. I brought up the security aspect because it is an argument which can actually appeal to those worried about loss of control and proliferating VMs. (I realize I probably have had a sheltered life, but I have only once been in a place that had more groups than people, with the groups controlling file access named so everyone knew what the main business process was and what the sub-task was.) --Ian Richmond -Original Message- From: Code for Libraries [mailto:CODE4LIB@LISTSERV.ND.EDU] On Behalf Of Genny Engel Sent: Monday, July 11, 2011 2:51 PM To: CODE4LIB@LISTSERV.ND.EDU Subject: Re: [CODE4LIB] Trends with virtualization I *had* the entire computer lab go down when the network failed once. That's when I switched it all to local desktops. The security was way easier to manage with a hosted desktop (I basically didn't have to manage it at all) but we weren't set up to offer any alternative when the network server hiccupped. It took me a lot of time to learn how to set up adequate security on an individual desktop, but once I got a good profile set up, I copied the image to all the other PCs and we were set. There weren't any equipment cost differences either way, as I recall. On moving things to the cloud, I'm still leery, especially after that Amazon thing a few months ago. http://www.dailymail.co.uk/sciencetech/article-1379474/Web-chaos-Amazon-cloud-failure-crashes-major-websites-Playstation-Network-goes-AGAIN.html Genny Engel Internet Librarian Sonoma County Library gen...@sonoma.lib.ca.us www.sonomalibrary.org 707 545-0831 x581 -Original Message- From: Code for Libraries [mailto:CODE4LIB@LISTSERV.ND.EDU] On Behalf Of Madrigal, Juan A Sent: Monday, July 11, 2011 8:21 AM To: CODE4LIB@LISTSERV.ND.EDU Subject: Re: [CODE4LIB] Trends with virtualization Its true what they say, history does repeat itself! I don't see how virtualization is much different from a dummy terminal connected to a mainframe. I'd hate to see an entire computer lab go down should the network fail. The only real promise is for making web development and server management easier. Vmware is looking to make thing easier with CloudFoundry http://cloudfoundry.org/ along with Activestate and Stackato http://www.activestate.com/cloud I definitely want to take those two out for a test run. Deployment looks dead simple. Juan Madrigal Web Developer Web and Emerging Technologies University of Miami Richter Library On 7/11/11 10:38 AM, Nate Vack
Re: [CODE4LIB] Trends with virtualization
As noted in my original message and in the responses, there's an argument to be made for desktop virtualization in the lab environment, assuming you don't have too many images to configure and maintain, you have enough devices to warrant the ROI, etc. (The environmental impact is definitely something to think about too.) My only comment back to Jeff is that your story would be more compelling if your email said, A year ago, we deployed... Desktop virtualization stories nearly always seem to be in the future tense. So you are hereby tasked to report back a year from now. Karen G. Schneider schnei...@hnu.edu
Re: [CODE4LIB] Trends with virtualization
On 7/15/11 9:14 AM, Karen Schneider kgschnei...@gmail.com wrote: My only comment back to Jeff is that your story would be more compelling if your email said, A year ago, we deployed... Desktop virtualization stories nearly always seem to be in the future tense. When I worked for Indiana University, their library IT group was considering going the thin-client route. This was roughly three years and I don't know whether they followed through with that or not. I think some of them are on this list and might be able to answer. I went virtual for servers several years ago and have never looked back. It's the best thing. Desktop virutalization has existed in the form of netbooting for a while. I saw a presentation five or so years ago about a university in Japan that had all of their labs running in a netboot environment. This was using Mac OS, however. I've also toyed with Ubuntu's LTSP [1] and was impressed with how easy it was to setup. It might not work for environments requiring specialized software that doesn't run under Linux, but for stations just needing web browsers for searching your catalog, it could be an option. The Ubuntu setup requires an ubuntu server to dish out the images, however I was able to run that under KVM, so a virtual server servering out virtual clients. That combined with clustering for your bare-metal servers would get you pretty close to 100% uptime. ...adam Adam Wead Systems and Digital Collections Librarian Rock and Roll Hall of Fame and Museum 216.515.1960 (t) 215.515.1964 (f) 1. https://help.ubuntu.com/community/UbuntuLTSP http://www.rockhall.com/ Rock Roll: (noun) African American slang dating back to the early 20th Century. In the early 1950s, the term came to be used to describe a new form of music, steeped in the blues, rhythm blues, country and gospel. Today, it refers to a wide variety of popular music -- frequently music with an edge and attitude, music with a good beat and --- often --- loud guitars.C 2005 Rock and Roll Hall of Fame and Museum. This communication is a confidential and proprietary business communication. It is intended solely for the use of the designated recipient(s). If this communication is received in error, please contact the sender and delete this communication.
Re: [CODE4LIB] Trends with virtualization
I think server virtualization is (mostly) a well-documented win, though I recall a scenario from several years back, in another organization, where too much was expected of one virtualized server. Regarding desktop virtualization -- the topic of this thread -- it appears Indiana is close to deployment: http://uits.iu.edu/page/azor K.G. Schneider schnei...@hnu.edu
Re: [CODE4LIB] Trends with virtualization
As a sometime lurker on this list (too many lists, too little time) I caught some of the conversation about virtualization. We have a small IT unit (5 staff including myself) and a very small VMWare server environment (9 windows servers, 1 linux) which we've been running since 2008. We're about to embark on desktop virtualization for our public computing areas (projected to be 54 workstations, currently 48 physical workstations). In some ways I'm going on a limb here as we don't know how well it will work until we do it. However , I hear nothing but good things about it from our sister library, which beat us to the punch, and has been using Citrix with more workstations than we will be doing for well over a year. 1) Server virtualization has been like a dream come true for us. Performance has been great, first off. Downtime was lessened from the day we went live, and since then we've gotten even better at maximizing virtualization's advantages in reducing downtime during individual server and software upgrades . Development is so much easier since we can clone production environments, as well as quickly and easily build test environments. My personal belief is that due to our backups of virtual disk images, we are way better off in terms of disaster preparedness than we ever were with just traditional backups. Is it a bit more expensive than traditional servers in our situation? probably. Is it worth whatever extra cost? For me, no question. 2) Desktop virtualization/thin client seems ideal for a public computing environment, where the big need is simply for web browsing and office applications. We don't think it will be cheap to do. However, it doesn't come off too badly. Even with the cost of the storage and server back end, it will cost well under the budget of a 3 or 4 year replacement cycle of traditional workstations. (I know, lots of us can't afford to do that. But...) The OS and virtual workstation per unit licensing is tricky and seems but for us, we're looking at a 3 year license of VMWare View, with no licensing cost beyond that for the Windows 7 OS. Thin or zero client devices have a much longer projected life than workstations. Lastly, energy savings can be significant in large environments. But even in smaller ones I would argue that buying fewer pieces of hardware - less plastic - and just smaller/fewer motherboards, is significantly better for the environment. On our current physical workstations, now out of warranty for a year, like most libraries we use a lockdown solution. We have workarounds and scheduled ways to patch the machines. We work with the campus IT and their LANDesk patching solution for afterhours scheduled patch times. However no matter what we do we still seem to end up touching each individual machine way more often than we'd like, with less than ideal results, amounting to what I think is a less than ideal customer experience. In virtualization we will have one image that we patch and correct as often as we want to, and it will always be exactly the same for our users no matter where they sit. Using either alternate VM profiles or ThinApp (we haven't decided) we can also license fewer copies of software (say, Endnote) and deliver just as much software as we need right to wherever the customer may choose to sit. Single point of failure? Sure - but for our library customers, if the Internet is down or the network is down, there's little to zero computing they would want to do anyway. So that's where we are, one year from now , perhaps I'll be singing a different tune, but we've thought about this and tested quite a lot before making the plunge. Jeff Kuntzman Head, Library IT jeff.kuntz...@ucdenver.edu 303.724.2126 http://hslibrary.ucdenver.edu University of Colorado, AMC Health Sciences Library
Re: [CODE4LIB] Trends with virtualization
Thanks, all -- most excellent commentary. You're a fount of common sense! Karen G. Schneider schnei...@hnu.edu
Re: [CODE4LIB] Trends with virtualization
On Sun, Jul 10, 2011 at 9:46 AM, Karen Schneider kgschnei...@gmail.com wrote: My down-home-country-librarian observation that I always tack on (with plenty of disclaimers) is If virtualization were the answer, we'd see more of it by now. This. Various vendors have been pushing the run all your desktops in the server room and export your I/O over ethernet solution for a long time. Heck, X11 does exactly this, and it's as old as the original Macintosh. I suspect the problems partly come down to the end-user experience (performance, customizability, etc) and partly the fact that making an environment truly truly homogeneous is not completely realistic in most environments. Once you've gone the everything will be virtualized route, making one desktop setup just a little different (adding custom hardware, etc) is nearly impossible. So it winds up making more sense to find a solution that lets you cost-effectively manage lots of desktops, because that solves your actual business needs, not what IT wishes your business needs were. That, and the fact that the parts of desktop hardware that usually fail tend to be the things people spend time touching with their dirty fingers and pouring their coffee on. Disks and motherboards do fail, but if you've done your homework right, you should be able to swap another one in within minutes -- and thin clients can fail, too. So virtualizing doesn't get you out of the business of heading out to replace gear. And desktop PCs are dead cheap and you can buy them from anyone. Custom virtual solutions usually want you to source from one vendor. That said: we do love virtualization for delivering Windows apps to Macs and Linux clients. Sometimes, there's just no substitute for SPSS on Windows. -n
Re: [CODE4LIB] Trends with virtualization
Its true what they say, history does repeat itself! I don't see how virtualization is much different from a dummy terminal connected to a mainframe. I'd hate to see an entire computer lab go down should the network fail. The only real promise is for making web development and server management easier. Vmware is looking to make thing easier with CloudFoundry http://cloudfoundry.org/ along with Activestate and Stackato http://www.activestate.com/cloud I definitely want to take those two out for a test run. Deployment looks dead simple. Juan Madrigal Web Developer Web and Emerging Technologies University of Miami Richter Library On 7/11/11 10:38 AM, Nate Vack njv...@wisc.edu wrote: On Sun, Jul 10, 2011 at 9:46 AM, Karen Schneider kgschnei...@gmail.com wrote: My down-home-country-librarian observation that I always tack on (with plenty of disclaimers) is If virtualization were the answer, we'd see more of it by now. This. Various vendors have been pushing the run all your desktops in the server room and export your I/O over ethernet solution for a long time. Heck, X11 does exactly this, and it's as old as the original Macintosh. I suspect the problems partly come down to the end-user experience (performance, customizability, etc) and partly the fact that making an environment truly truly homogeneous is not completely realistic in most environments. Once you've gone the everything will be virtualized route, making one desktop setup just a little different (adding custom hardware, etc) is nearly impossible. So it winds up making more sense to find a solution that lets you cost-effectively manage lots of desktops, because that solves your actual business needs, not what IT wishes your business needs were. That, and the fact that the parts of desktop hardware that usually fail tend to be the things people spend time touching with their dirty fingers and pouring their coffee on. Disks and motherboards do fail, but if you've done your homework right, you should be able to swap another one in within minutes -- and thin clients can fail, too. So virtualizing doesn't get you out of the business of heading out to replace gear. And desktop PCs are dead cheap and you can buy them from anyone. Custom virtual solutions usually want you to source from one vendor. That said: we do love virtualization for delivering Windows apps to Macs and Linux clients. Sometimes, there's just no substitute for SPSS on Windows. -n
Re: [CODE4LIB] Trends with virtualization
As others have mentioned, you have to know what problem you're trying to solve before you start looking at solutions. Crazy, I know... We're looking at a virtual lab environment because we don't have a lot of qualified lab managers in departments that refuse to allow central IT to manage their labs. Long story, politics, etc... But for this scenario we can provide custom images with the right software and better access all the while freeing up a room. Basically the student can access the lab image from their own machine or any other machine without needed to be in a physical lab. This is great if you don't need any special hardware in the lab. Also, as previously mentioned, if you can't cover 100% of the desktops what is the cost going to be to have two ways of managing desktops and will the savings in going virtual cover the added cost of doing the same thing two ways? Exceptions cost money and seem to rarely be factored into the final bill. Server virtualization has been wonderful though. Savings in electricity are real[1]. Physical servers are still called in when needed (Oracle, etc). If you want to start somewhere, this is the place. Pat [1] Sorry, I don't have the numbers but I've been told they're real. On Sun, Jul 10, 2011 at 7:46 AM, Karen Schneider kgschnei...@gmail.comwrote: Dear C4L community, One of the VPs on campus asks me from time to time on trends with virtualization in academic settings -- specifically, virtualized desktops. My own response (qualified with I am not an IT person, but...) has been that I believe, based on what I read, that this highly-promising technology isn't more widespread for several interrelated reasons (that are also applicable to our campus environment: a) ROI is not as clear, especially in smaller environments (startup cost, network, storage); b) university WANs are often not be robust enough to support virtualized desktops (and I'd add, we're on an uphill Sisyphean climb with bandwidth--there will never be enough of it); c) outside of the lab/classroom environment (where I think an argument can be made for virtualization, if other conditions are met, and the campus has the expertise to deploy/manage this environment), the ROI of a virtualized desktop may be mooted by the need for individualized desktops; d) it's a single point of failure. My down-home-country-librarian observation that I always tack on (with plenty of disclaimers) is If virtualization were the answer, we'd see more of it by now. I realize that's a humble insight, but given how many talks I've been to over the past decade about what virtualization *would* be doing, versus what it *has* done, I think it's not entirely invalid. I also pointed the Veep toward this article: http://www.networkworld.com/news/2009/061809-desktop-virtualization.html So... any thoughts? Resources? POVs? Etc.? (If you want more context for this inquiry, write me off-list.) Thanks, dear old C4L community-- Karen G. Schneider Director for Library Services Holy Names University http://library.hnu.edu schnei...@hnu.edu
Re: [CODE4LIB] Trends with virtualization
We are using Xen server virtualization in our data center, and we like it. We use the licensed version from Citrix in order to get professional support. This qualifies us to receive the gamut of Citrix virtualization marketing materials and invitations to Citrix events. Amusingly, none of these have anything to do with the products we use. They are entirely devoted to Xen desktop and application virtualization products. It is clear that Citrix bought Xensource from its founder and the University of Cambridge with the full intention of using it to create these products. Why? In the words of famed bank robber Willy Sutton, like banks, It's where they keep the money. So far, I have yet to see anything other than marketing fluff to support the argument that commercial desktop virtualization products are cost effective. The big difference between server virtualization and the kind of desktop virtualization we are discussing (as opposed to, say, Parallels) is that server virtualization requires not additional hardware. The real savings, according to the sales folks, comes from IT operations. At least that is the theory. As you mention, we have seen many iterations on this theme from dumb terminals and serial distribution to virtualization, and these all depend on a belief that those solutions are cost effective. Were there not existing, cheaper ways of remotely managing desktops and desktop applications that work on inexpensive commodity hardware, it might be easier to make that argument. Thanks, Cary On Sun, Jul 10, 2011 at 7:46 AM, Karen Schneider kgschnei...@gmail.com wrote: Dear C4L community, One of the VPs on campus asks me from time to time on trends with virtualization in academic settings -- specifically, virtualized desktops. My own response (qualified with I am not an IT person, but...) has been that I believe, based on what I read, that this highly-promising technology isn't more widespread for several interrelated reasons (that are also applicable to our campus environment: a) ROI is not as clear, especially in smaller environments (startup cost, network, storage); b) university WANs are often not be robust enough to support virtualized desktops (and I'd add, we're on an uphill Sisyphean climb with bandwidth--there will never be enough of it); c) outside of the lab/classroom environment (where I think an argument can be made for virtualization, if other conditions are met, and the campus has the expertise to deploy/manage this environment), the ROI of a virtualized desktop may be mooted by the need for individualized desktops; d) it's a single point of failure. My down-home-country-librarian observation that I always tack on (with plenty of disclaimers) is If virtualization were the answer, we'd see more of it by now. I realize that's a humble insight, but given how many talks I've been to over the past decade about what virtualization *would* be doing, versus what it *has* done, I think it's not entirely invalid. I also pointed the Veep toward this article: http://www.networkworld.com/news/2009/061809-desktop-virtualization.html So... any thoughts? Resources? POVs? Etc.? (If you want more context for this inquiry, write me off-list.) Thanks, dear old C4L community-- Karen G. Schneider Director for Library Services Holy Names University http://library.hnu.edu schnei...@hnu.edu -- Cary Gordon The Cherry Hill Company http://chillco.com
Re: [CODE4LIB] Trends with virtualization
On Jul 11, 2011, at 11:21 AM, Madrigal, Juan A wrote: Its true what they say, history does repeat itself! I don't see how virtualization is much different from a dummy terminal connected to a mainframe. I'd hate to see an entire computer lab go down should the network fail. The only real promise is for making web development and server management easier. re: web development I assume by that you're talking about cases like Citrix, where they force you to come in from the same OS web browser version, so they don't have to worry about Firefox rendering differently from Safari, or the IE6 vs. 7, etc. It's okay for an intranet, but I don't know that it's a good idea for general web usage, as they normally force people to use some outdated browser, as the web applications always seem to be designed for IE6, and never tested on anything else. (if they were, they then try to serve down alternative versions using browser detection, which in my experience is more likely to make things worse) ... The only reason I've heard to virtualize desktops wasn't for monetary considerations, and wasn't for general word processing and such ... it was for workstations for scientific processing. By using virtualized servers, you can more easily take snapshots of the machine's state to archive it, and later restore it to re-run the software. This gives you two advantages: (1) reduced down-time for patching / upgrading software -- you patch the image, then push the image into the processing pipeline. (2) Because you've archived the OS, libraries and all software, you have something you can analyze should someone identify problems with the data processing such as discontinuities after an update. I could see the first one being useful for most groups, but with tools like puppet and chef, it might not be a big deal. I can't remember what the software was that the university I formerly worked for used in their computer labs -- it basically reset the machine on each login, in hopes to prevent someone from installing malware (intentionally or accidentally) that would then affect later users. And then once a week each lab was closed down so they could do a complete re-format and re-image of each machine ... you might be able to do something similar with virtual desktops. -Joe
Re: [CODE4LIB] Trends with virtualization
I *had* the entire computer lab go down when the network failed once. That's when I switched it all to local desktops. The security was way easier to manage with a hosted desktop (I basically didn't have to manage it at all) but we weren't set up to offer any alternative when the network server hiccupped. It took me a lot of time to learn how to set up adequate security on an individual desktop, but once I got a good profile set up, I copied the image to all the other PCs and we were set. There weren't any equipment cost differences either way, as I recall. On moving things to the cloud, I'm still leery, especially after that Amazon thing a few months ago. http://www.dailymail.co.uk/sciencetech/article-1379474/Web-chaos-Amazon-cloud-failure-crashes-major-websites-Playstation-Network-goes-AGAIN.html Genny Engel Internet Librarian Sonoma County Library gen...@sonoma.lib.ca.us www.sonomalibrary.org 707 545-0831 x581 -Original Message- From: Code for Libraries [mailto:CODE4LIB@LISTSERV.ND.EDU] On Behalf Of Madrigal, Juan A Sent: Monday, July 11, 2011 8:21 AM To: CODE4LIB@LISTSERV.ND.EDU Subject: Re: [CODE4LIB] Trends with virtualization Its true what they say, history does repeat itself! I don't see how virtualization is much different from a dummy terminal connected to a mainframe. I'd hate to see an entire computer lab go down should the network fail. The only real promise is for making web development and server management easier. Vmware is looking to make thing easier with CloudFoundry http://cloudfoundry.org/ along with Activestate and Stackato http://www.activestate.com/cloud I definitely want to take those two out for a test run. Deployment looks dead simple. Juan Madrigal Web Developer Web and Emerging Technologies University of Miami Richter Library On 7/11/11 10:38 AM, Nate Vack njv...@wisc.edu wrote: On Sun, Jul 10, 2011 at 9:46 AM, Karen Schneider kgschnei...@gmail.com wrote: My down-home-country-librarian observation that I always tack on (with plenty of disclaimers) is If virtualization were the answer, we'd see more of it by now. This. Various vendors have been pushing the run all your desktops in the server room and export your I/O over ethernet solution for a long time. Heck, X11 does exactly this, and it's as old as the original Macintosh. I suspect the problems partly come down to the end-user experience (performance, customizability, etc) and partly the fact that making an environment truly truly homogeneous is not completely realistic in most environments. Once you've gone the everything will be virtualized route, making one desktop setup just a little different (adding custom hardware, etc) is nearly impossible. So it winds up making more sense to find a solution that lets you cost-effectively manage lots of desktops, because that solves your actual business needs, not what IT wishes your business needs were. That, and the fact that the parts of desktop hardware that usually fail tend to be the things people spend time touching with their dirty fingers and pouring their coffee on. Disks and motherboards do fail, but if you've done your homework right, you should be able to swap another one in within minutes -- and thin clients can fail, too. So virtualizing doesn't get you out of the business of heading out to replace gear. And desktop PCs are dead cheap and you can buy them from anyone. Custom virtual solutions usually want you to source from one vendor. That said: we do love virtualization for delivering Windows apps to Macs and Linux clients. Sometimes, there's just no substitute for SPSS on Windows. -n
