admittedly - the software would have to support unattended instllation, meaning you provide all the instructins either in a config file or as command line arguments. but i have found almost all software for windows systems supports unatended installation these days
you know how one set of updates install - a reboot is required and then sometimes another set of updates are needed. you can use poweshell to keep installing the updates even when a reboot is required until there r no more to install
powershell has a very good restart command in version three which allows you to perfomrm a reboot, wait for the machine to come back up and then carry on with your script.
Well I had to, I would always have to walk away during the long parts of the installation and guesstimate when to walk back into the datacenter to continue
seemed logical to have it alert me when all was finished so no time was wasted
Role dependent, I know I will get body slammed in this room for not using an image, but the infrequency in which we need to install workstations has swayed me from that. I do it the old school way
Whoever their manager is going to be gives me a list of software they will need, assigns it to me in our internal work request system, and I do it
I've never bothered much with imaging. Mostly because we have PCs of all different models. I use PXE (Pre-Execution Environment). The Windows setup is 100% automated as is the installation of all the software and lastly the windows updates.
There is going to be a dark week in the near future where I will have to replace something in the neighborhood of 50 windows boxes in our datacenter with the cute little poweredge R210IIs
we were using towers for a particular set of servers back in the day, all installed with windows 7 and serving a single purpose, in a distributed fashion
My boss probably has his own reason, but I am assuming that he does it for fault tolerance or something
If one physical server dies, it's not hosting anything but whatever application it is desinged to serve
And he explained something to be before that I was a little to inept to absorb at the time, but I remember him saying something about horizontal scaling vs virtical scalig
So as we expand, we can simply add more devices to handle the distrubuted workload, rather than continue to upgrade one or two mega servers to handle the increase
Well, when a single piece of hardware fails (which you'll have fewer pieces of hardware, so fewer parts to potentially fail), your virtual machines with application workloads will be automatically rebooted on a working piece of hardware instead of staying down.
The windows boxes I mentioned all exists soley to run a program, this program runs an instance of itself on each CPU core, through some programatic witchcraft
the program uses each core seperately to process incoming files, that are randomly selected from a central server, thus distributed
So with 50 some odd windows boxes with 4 cores per CPU, to try and stick all of those on one server and actually have the resources ( 50 X 4 cores each) would be very difficult at best
I am assuming that VM's let you allocate cores and other resources on install time
@Miguel No need for you to run it all one one server; you'd essentially have a cluster of physical hosts running your virtual machines distributed between them.
@Miguel If you overcommit (assign more virtual cores than there exist physical cores to run them on), then there can be contention between the virtual machines for CPU time, yes.
@Miguel Right - and a modern virtual machine host might have something like 4 CPU sockets, each with a CPU that has 10 cores with hyperthreading - lots and lots of execution threads available.
my point being - that the utilisation of each core by your program might be miniscule. So once you make it multithreaded you might actually get away with one core...
Yeah, it depends a lot on the workload. If they ever back off their CPU usage (like when waiting for a new file?), then you might be able to allocate more vCPUs than you have physical cores in the system.
unfortunately - in IT, as soon as something new comes out, everyone turns nasty on the old way of doing things. Sometimes we have the sense to go back though.
The searching is for OCR. So the documents come in and are initially stored on one of a series of NAS servers. A copy of the document is sent to one of the windows boxes I talked about to get the text of the documents parsed and associated with the file by an ID. Once this is done the results get sent to a large linux db server which has the information on every document, the text associated with each document ( for text searching) and the location of the actual document on the NAS servers.
When somebody searches the system for text within a document, rather than massive amounts of queries htting that one linux servers, we hand a section of the results to one of the other linux servers. So now when somebody looks for text on a document, the results can come from one of 60 servers, which each server gets enough queries a day to make 24 GBs of RAM just barely enough to return the results within subseconds.
A database shard is a horizontal partition in a database or search engine. Each individual partition is referred to as a shard or database shard.
Database architecture
Horizontal partitioning is a database design principle whereby rows of a database table are held separately, rather than being split into columns (which is what normalization and vertical partitioning do, to differing extents). Each partition forms part of a shard, which may in turn be located on a separate database server or physical location.
There are numerous advantages to this partitioning approach. Since the tab...
I'm not a database guy so I don't know it off the top of my head
As of right now I don't know anything other than the fact that it seems to be all the rage. But with minimal research it looks as if it may be a good way to cut some costs and free up some real estate in the datacenter
I think I just like my virtualization solution to be a slam-dunk... straight-ahead. I don't have to worry about hypervisor stability on the VMware side. The kernel is refined well enough, supports the hardware I need... Good consideration has been given so I can PXEboot hosts or run off of SD card. Performance always improves with each revision.
@dawud that's just the start of what we need :) It needs to work with our asset mgmt system, ldap role based auth andnot want control of things like dns and monitoring
most 'server management' or 'vm management' software out there wants to control your entire environment, which is annoying to the point of not being usable.
For instance, I want just one and exactly one graphing system, so I can easily correlate between any two metrics. So don't make your vm management thing do its own monitoring and graphing for crying out loud.
Being able to plot e.g. pageviews per minute and packet drops on memcached servers in the same graph is an insanely useful debugging aid (in this case it pointed out that the packet drops started well before the drop in pageviews, making it an unlikely cause of that drop)
@dawud we use diamond for symple measuring and the socket interface to pipe lots of custom metrics from our business monitoring system to it
Environment:
- Storage: HP P2000 MSA G3 SAS Array with 24 300GB 10k SAS Disks
-- Two Storage Controllers with redundant SAS connections to each host
- Hosts: Three HP DL380 G7s with a 10GB SD Card, CPU, Ram, etc...
ESXi 5.0 is installed on the SD card in each host, this is the only local sto...
Desired result is ability to survive many disk failures. RAID6 gives me the ability to lose 2 disks per vDisk and still be okay. Applications that the servers run cannot go down. This is most important. Performance isn't much of an issue because the system is small. Right now I'm using Storage vMotion to move VMs off of a LUN with a bad disk, so that disk reconstruction performance doesn't affect performance. — Lucretius24 mins ago
didn't read the question. But fuck raid 5/6. 1+0 and spares is enough. I have lots of disks die because I have lots of disks. Rarely more than one in the same server and so far never lost a box due to too many disks failed.
did lose a box due to it being raid 0 and a disk got pulled because it said 'predicted failure'. That's just human stupidity though.
@ewwhite that RAID61 guy works for Kroger. He really shouldn't have problems getting the budget for doing his service properly if it's all that critical to operations.
@Adrian My produce clients deal with Kroger... I've had to CALL Kroger's IT staff multiple times and walk them through Microsoft Exchange server settings because they would inadvertently do something that caused mail rejections.
@ewwhite Ah. And there's days that it really doesn't pay to live in Seattle. Too much talent here to compete with if guys like those can hold a job down. Too bad I'd never want to live in Atlanta.