The short summary is that you are using your IBM i system as a SAN for
the windoze servers.
As to the size and or form factor (blade vs pizza box) you pick what
works for your company. If you only need two or possibly 3 servers then
pizza boxes will probably work best. If you see yourself with 6, 8, 12
servers then blade center has lots of advantages. I won't go into all
that here.
For the integration pieces consider these factors:
Obviously Power System disk subsystems are robust and reliable but so
are many SANs today.
IBM i provides up to 10GbE connections for iSCSI and with IBM i TR3 you
can bind them together for even more bandwidth and redundancy. Many SANs
also support this.
IBM i allows you to create 'disks' (objects in the iFS actually) for the
windows servers. These can be small to large as needed, can be created
and allocated to the servers on the fly. Give the servers only what they
need capacity wise because IBM i has already spread the disk images
across many physical disks. Protection is also afforded by IBM i so the
servers 'think' the disks are unprotected. Most SANs also support this.
IBM i allows you to link the disk to multiple servers so that you can
support things like VMWare's VMotion capability to move workload between
servers. Most SANs also support this.
So if most SANs also support this why i? I knew you'd ask this :-)
You already know how to run i. You already have an i. You can add disk
to your i with no additional license cost. I don't have to tell you how
reliable i on POWER is. Dollar per TB on i vs most SANs WITH COMPARABLE
CAPABILITIES (i.e. similar sized disks and I/O cards.) is about the same
spend.
So why by a SAN if you already have i?
I has good backup capabilities built in. With i 6.1 and newer you can
back up the servers disks on the fly and on many systems today a single
LTO5 tape can hold the entire system. This includes all of i, and all of
your servers. There is no better way to back up your entire
infrastructure than this. There just isn't. In a DR scenario you walk in
with one tape that gets your entire server infrastructure back to a
known point with one tape. Add daily saves and you're good to go. I
won't even try to describe doing this on a SAN because, well, SANs don't
have tape drives.
So why by a SAN if you already have i?
You can clone disks for easy server deployment on i. Just CRTNWSSTG and
specify the space to clone. You can do this on some SANs as well.
You can WRKNWSD and vary off a server that isn't behaving and vary it
back on to reboot it. You cannot do that with SAN. With SAN you must
connect to the service processor of the server in question and using
that servers interface power it off and then back on. Is that hard? Not
so much but it's a different interface on a different IP with a
different user and password to remember. WRKNWSD is just there on your
command line. Oh right, it's also in iNav if you're a fan. Sure Windoze
has gotten more reliable over the years but this alone used to be one of
my favorite features!
'Improved uptime through consistent deployments', really? Yes. Really! I
have to laugh (because otherwise I would cry) when I walk into shops
with 'server of the day' all over their racks. THis one has a quirky
RAID card, that one odd format disks, over here the CD drive is cranky
and they use a USB unit. That on has a freeked out video card. The
fourth one down has a failed Ethernet (port 1 only) but the driver blows
up if we don't have a cable in it anyway. You think I jest but I've seen
all this crap. Biggest problem of course is moving the workload off said
servers to get them fixed or just plain retired and that's because you
can't simply pull the disks and stuff them into another server, at least
hardly ever. With all this integration stuff we simply re-assign the
disks to another server and crank it up. IBM says something like 10
minutes to do this but in reality if you know your servers it can be 2
min or less. Whether you're doing this for maintenance or failure it's
way cool stuff.
So why by a SAN if you already have i?
SANs are not some awful, bad _e_vil _m_achines _c_ompromising our data
centers. However if you don't have one yet, it IS yet another piece of
computing to learn, to purchase, to license, to maintain, and yes to
figure out how the dickens you are going to back it up. Anyone want to
buy a 50TB DS4800 from a customer who's learning all this the hard way??
By the way, Carl just mentioned his iSCSI setup. You want to learn how
to minimize the server footprint in your data center? There's your guy!
- Larry "DrFranken" Bolhuis
On 11/10/2011 5:02 PM, Kurt Anderson wrote:
Hey all,
I posted earlier about running jspwiki on the i, and this has really snowballed. The wiki we're getting up and going on the i, however I had the bright idea of seeing about getting all of our servers here running as reliably as our Power system, and now I'm tasked to investigate. My expertise is RPG, and while I have some system knowledge, that is definitely not my forte. My train of thought was to get our other servers onto a Power box, although what I'm finding is slightly different.
Currently the software we develop for Windows uses SQLServer (2005 and 2008). To switch to another database would be an immense undertaking, so we're currently working with the idea of keeping SQLServer for that software.
I had thought that the IBM i could be partitioned (if that's the right word) to run the Windows operating system, but everything I'm reading in regard to the Power system's support for Windows is to purchase either System x Servers or a IBM Blade Center and to connect via iSCSI.
It sounds like there is no real performance increase in doing this (according to the faq). Presumably the cross platform (SQLServer to IBM i) data access will be faster since they'll be connected via a gigabit line - although we don't do much of that today, it's nice to know in case that changes.
The big advertised benefits seem to be (from their brochure):
1. Flexible Server Deployment
a. The 'hot spare' sounds like a really slick server recovery method
2. Simplified Storage Management
a. Easily and dynamically assign disk storage for each server. Less 'wasted' hard drive space per server.
3. Synchronized Security
a. I guess I do like the idea of having the same sign-on to all servers, not a big selling point.
4. Innovative Integration
a. Reducing administration and maintenance costs is nice, though I need to research our end to see what kind of savings we'd expect
b. "Run applications you need using resources and skills already in place" - except we need to buy new System x Servers or Bladecenter servers - I'm sure the cost of these varies with the number of cores and memory, but I really have no clue if these are in general cheaper or more expensive than a comparable windows server.
5. Streamlined Communications
a. I do like the 'fewer points of potential failure'
6. Improves Windows server uptime through consistent implementations
Given all of that, I'm wondering what people's experiences are with iSCSI. Are there any benefits you see/saw that I don't have listed. Are there things I've listed that you don't see as much of a benefit?
Is the only difference between the Blade and the System x the size of the hardware?
Random fyi: I do have the IBM I iSCSI Solution Guide, so I've been perusing that as well.
Btw - I find it funny that the course number for System I Integration with BladeCenter and System x is AS300 (not quite AS400).
I appreciate your input,
Kurt Anderson
Sr. Programmer/Analyst
CustomCall Data Systems
midrange.com
