What is "bare metal provisioning"?
"Bare Metal" provisioning refers to the process of installing an operating system (and often applications, data, configurations, scripts) onto a system that doesn't have an operating system installed on its hard disk. It can also mean over-writing a hard disk with a different operating system (and other files), in which case you lose everything that had been installed on the drive previously.
- Using CD-ROMs or a DVD, one system at a time. This works fine if you have fewer than a handful of systems and you don't re-purpose systems often.
- Using "home grown" tools and scripts that allow systems to be provisioned over a Local Area Network. These traditionally are very complex to set up, and the knowledge of how to use them resides in the head of only a few people.
- Using commercial products that are part of broader systems management solutions used in large enterprises (sophisticated, policy-based, expensive, often hard to justify financially)
- Using LinMin Bare Metal Provisioning, whereby you upload all the Linux, Windows and ESX/ESXi ISO files (or DVDs) for all the operating systems and hypervisors you use only once (benefit over solution 1), then specify how and when systems get provisioned and with what OS/hypervisor.
- Any system that has PXE boot capability (essentially all Intel and AMD x86 and x86-64 systems today) that is connected to a local area network.
- Servers (1u, 2u, 4u, etc), blades, appliances, desktops and laptops, with single or multi-core CPUs, with one or more CPUs. So just about any type of system used in your IT department.
- Also you can provision VMware virtual machines as well as physical systems. This is very useful to build the initial VMs before you clone them. It also lets you provision physical systems and VMware systems without changing anything. With LinMin, physical and virtual systems are interchangeable
- Maybe you should. If you have a highly complex IT environment, and you want to have on-demand "utility computing" where systems get repurposed automatically based on dynamic load, and if you have the funds to justify this purchase, then this may be the way to go for you. However, with LinMin's Application Programming Interface (API) in release 5.2 your existing systems management tools can trigger provisioning events trough LinMin.
- If you don't want to have to go up 4 levels of company management to justify the purchase of tens of thousands of dollars (and instead pay only a few hundred), and you're satisfied with an easy to use solution that gives you the flexibility and control over how your systems get provisioned (and that didn't take you months to write yourself), then LinMin is the better choice.
- And you can still automatically install your Tivoli or BMC or other management Agent so the system will "fit right in" (be ready to be managed by your systems management software) when it's ready.
- We'll admit it can be a fun and tempting project! But consider the time involved to build the environment, document it, test it, and get management buy-in. This will take weeks of dedicated research, scripting, networking and testing. Then consider the on-going maintenance efforts, getting calls while/if you're on vacation because something doesn't work, etc.
- In contrast, for a few hundred dollars, you get from LinMin a fully supported, easy to use product that installs in a few minutes with a single command (which in itself is pretty amazing considering that LinMin downloads, installs and configures your database, network services, etc.) It has a friendly UI, scripts to help upload your distros/OS, ability to preconfigure systems and keep these configurations ready to go (e.g., Red Hat Enterprise Linux Advanced Server, RHEL Advanced Server configured as a web server, RHEL Enterprise Server with an email system pre-configured, Windows 2003 Server, ESXi, etc.) We recommend you apply your domain expertise on what your system should look like when ready, rather than re-invent the wheel.
I have so many different hardware configurations in my shop (data center, lab, school, company) that I can never keep track of what hardware is on what systems. How does LinMin's products address this?
Easily. We don't assume we know the specific hardware ahead of time. When a legitimate PXE boot event occurs, we download on the system a mini-kernel that then grabs and executes the OS Vendor's hardware detection and system installation startup software (kickstart for Red Hat-based systems, YAST for Novell/SUSE-based ones, etc.) You can literally have 20 different systems side by side, and each one will get the correct drivers without you doing any manual intervention.
- Well, so you have to pay a few hundred dollars...
- You have to be a reasonably skilled system administrator and know what DHCP is, what a subnet is, what range of IP addresses you have, maybe open firewall ports depending on your security infrastructure, and decide what distros and control files you want to have installed on new or existing systems. Once you, the expert, has done this, then less skilled IT staff can seemingly perform miracles leveraging what you (the domain expert) have done, such as "get those 20 new servers off the truck, bolt them in the racks, plug them into the network, power them up, and make have of them Novell SUSE 10 database servers for Project A, and the other have RHEL or CentOS Application Servers with jBoss for Project B. Call me in 1-2 hours when you're done."
Just doing some simple math, considering my alternative is to spend hours stuffing CDs into systems, then more hours configuring them, a few hundred dollars seems pretty darn reasonable! Now even though it's a great value, what else can I tell my boss to get him or her to use a credit card to order this today?
- Think of the time you'll save (so you can do real work, complete other projects).
- Reduce the elapsed real time and your personal elapsed time ("labor cost, to put it bluntly") to get new systems running.
- Think about how fewer errors will be made when someone can simply click from a list of known-good systems/configurations vs. doing things from memory.
- Oh, and we didn't tell you yet: you can be in front of a new system, and you'll be presented with a list of distros/OS/configurations to select from, hit a number and walk away ("on-the-fly" provisioning). You can also pre-determine what system, based on its MAC address ("Fire and Forget" provisioning), will get what OS/distro and what applications, what IP address, root password and on and on.
- You can even capture entire disk images for a system, and copy them to another system on the network, then restore that image to bring a system to its "last known good" state.
I'm an appliance vendor or a systems integrator: why should I use your product?
So your technicians can select on the manufacturing line what should be installed prior to shipping to the Customer "Is this a Firewall or Anti-Virus appliance? Does this University want CentOS or Fedora or Ubuntu pre-installed on these Intel servers?" LinMin has special pricing for system builders like you (where you provision a system, ship it, and never see the system again.)
How do you offer automated provisioning at those price points?
LinMin benefited from an acquisition of IP from a systems management company that went under. We leveraged years of product development and don't have to offer a product that reflects its true development costs. If we did, we'd probably charge $30-$50,000. The product requires very little tech support (single command install, and easy to use GUI), so we don't need a small army for tech support. Internet sales model (no expensive sales reps, company cars, fancy lunch expenses, etc) further cuts costs and helps profitability.
LinMin reduces energy costs by decommissioning existing hardware to save electric power and to cut the carbon emissions foot print? How so?
Because the product does both native, unattended OS installation (the equivalent of putting in DVD and answering a bunch of questions, and the OS detecting the hardware to install the proper drivers, etc.) as well as disk imaging (capture the contents of disks and store them on the network), we make it easy for customers to park systems (after capturing their disk contents), then natively provision new systems, install the apps, etc (then capture the disk image after for disaster recovery). So you image older systems (just in case you need to bring them back to life later), build "roles" for systems (e.g., RHEL JBoss, SLES 10 SP2 SAP server) and when the domain experts have the "recipe" (role) right, they can easily deploy the equivalent apps to newer hardware, then after a period of running in parallel, cycle down the power hungry systems.
Can I save labor costs by automating provisioning of Linux, Windows and ESX/ESXi systems through a single interface?
Yes, and by making it easy to do so. We typically see IT teams with specializations ("the Windows guys" vs. "the Linux guys"). With LinMin, the respective domain experts create and test the provisioning roles, then just about anyone can then provision (or image) systems without really knowing what they're doing ("hey, go install the role Windows 2003 Web Edition on systems x through y).
How does Fire and Forget work? Does this simply mean it enables so much automation that one person can do the provisioning of many systems?
Yes, you have a database of resources (your pool of servers or blades) on which you can pre-specify what will happen to them the next time they do a network boot. "Next time MAC address xyz boots to the network, capture its disk image; next time system abc net-boots, install CentOS; next time system 123 net-boots, do nothing, ever, under any circumstance) etc.
What are examples of deploying or re-purposing tasks where automating those through LinMin would reduce human error?
Because the provisioning roles are repeatable and refined until "just right" (install RHEL 5.2, install MySQL and PHP, run these configuration scripts, assign yourself this root password and these network configurations, install your whatever management agent, Tivoli, OpenView, whatever), the domain experts do their thing, then anyone can then assign that role to a given system, knowing there's a 100% chance of repeatability. If something isn't right, tweak the recipe (role).
How could LinMin impact web hosting companies with hundreds of systems?
By making it very easy to repurpose (or recover) systems. Most hosting companies offer Linux and Windows systems (many also use ESX/ESXi and other hypervisors). But they never know the right mix, so they can't really pre-stage them with a given OS without wasting time (or servers without the right OS). With LinMin, they can look at the resource pool, and deploy the right OS with the right apps (e.g., VirtualMin to have multiple domains reside on a single physical server). And with LinMin's API (Application Programming Interface), your other IT application software can tell LinMin what to provision, automatically.
How does LinMin help dedicated server companies with hundreds of systems?
Pretty much the same as above, except even better, because they don't need to install "multi-tenant" software. You want CentOS, RHEL, SLES or Windows? Voila, 15 minutes later, a machine from a spare pool is ready to roll and have the controls passed to the customer.
How can LinMin help SaaS companies?
Here's it's more a question of replication of systems in a traditional 3 tier architecture (DB, App, UI). They need to roll systems out like cookies, so if the hardware is identical, just slam an image on a box (cloning) and you've added a DB, App or Web server to the pool (and your load balancing HW/SW takes over). Or if you have different hardware, instead of cloning, you natively provision the system(s) so the OS does HW detection and loads all the right drivers).
What other companies could benefit from using LinMin?
Some companies have home-grown, open source PXE servers that can do the job, but typically only for 1 or 2 OS types, and if the 1 or 2 people with the knowledge in their heads leave on vacation or the company, it's panic time (it's all script based). For those companies, LinMin can do the job, only better, for less money than the labor it took to get the home grown system installed in the first place (LinMin is cheaper than Open Source, once labor costs are factored in). For companies that have an existing systems management framework, but don't want to spend hundreds of dollars for provisioning software per system each time a new one rolls into the warehouse then the data center, we're an alternative. Use LinMin to install the OS and the management agent, then we disappear until asked to do it again.
How does the API enable customers to add the automation capabilities of LinMin to their existing IT infrastructure?
This is really a neat feature. With an API, you can have other systems management (or customer management, or billing, or monitoring) applications issue "orders" to LinMin to provision systems. E.g, a dedicated server company can add logic to their integrated billing system, so when a customer fills out a form, asks for a certain OS, and enters the credit card info, a message gets sent to LinMin that automatically sets up a box to be provisioned to the customer specs ("deploy a bare RHEL server with this IP address, etc")
How does LBMP keep capital purchases under control?
For one thing, our software is so inexpensive that if falls under the category of expenses rather than capital budget. It also allows you to easily repurpose systems on the fly, so you don't have to have 50 DB servers (with 10 backup systems) and 50 App servers (with 10 backups). If a box goes dead, replace it while it's being fixed, and in the mean time go to your resource pool (of 10, not 20) systems and deploy the role you need.
How is LBMP great for facilitating cloud computing? By helping cloud computing services provision their customers' applications in the service provider's network / Internet cloud?
Yes, precisely. The trial and error approach to making the right Fondue (recipe, role) can be honed by the customer in the privacy of their data center, then when it's right, if they "go cloud", the simply upload LinMin to a cloud box and have it provision other systems. Also, the Cloud "owners" can use LinMin to have predictable, cost-effective systems available to cloud customers at the drop of a hat.