TSM can't compete

ChrisRees

ADSM.ORG Senior Member
Joined
Jul 24, 2008
Messages
542
Reaction score
13
Points
0
Location
Birmingham, UK
PREDATAR Control23

Just wanted to vent frustration at the high prices IBM come up with for Extended edition licenses with their PVU model...

I've got a requirement to backup 2 x 24 core intel servers and the PVU calculator = 4800 PVUs... just got a quote back and before discount its around £19k :eek: ... just to backup two servers, which seems completely ridiculous.

My Arcserve colleagues are laughing , they claim they can back them up for less than £100.. no way I can argue with management about which product suite should be used for these servers which are development environment.

Rant over
 
PREDATAR Control23

I know the pain and understand your frustration. We are actually looking for other alternative to backup our dev env. Cause TSM cost a arm and a leg.

Some of our prod server are not even 2% of the data we backup but wee need CPU and IO power.. so for a backup of not event 100gb Total with an 5gb daily change.. it's on a 4 CPU intel E6540 (6 Core per CPU) They are charging 100 PVU per core.. NOT the 50 PVU!!!!

i'm just happy they didnt buy anothe CPU.. They would charge 120 PVU per CORE!!!

more 14K just of 5GB of daily backuP!!!
 
PREDATAR Control23

Just wanted to vent frustration at the high prices IBM come up with for Extended edition licenses with their PVU model...

I've got a requirement to backup 2 x 24 core intel servers and the PVU calculator = 4800 PVUs... just got a quote back and before discount its around £19k :eek: ... just to backup two servers, which seems completely ridiculous.

My Arcserve colleagues are laughing , they claim they can back them up for less than £100.. no way I can argue with management about which product suite should be used for these servers which are development environment.

Rant over

Just a thought, but one way to consolidate license is to use a file server type mechanism to reduce licensing. In other words, backup your data from limited SAN/NAS appliances (file/data server appliances) versus per server.
 
PREDATAR Control23

I know the pain and understand your frustration. We are actually looking for other alternative to backup our dev env. Cause TSM cost a arm and a leg.

Amanda (zmanda) for development environment? We're not backing up dev stuff (well, only dev oracle db stuff anyway).
 
PREDATAR Control23

Mikeatkc. Interesting software... but dont know if it can handle 6TB a daily backup for a dev env.. Our dev env is about 500 Nodes with 6TB of daily backup

our prod is 700 nodes with 11TB of daily backup.

I would add that all the backup DEV/Prod our on the same VTL and 3584. Management is aware that if we add another backup software.. they will have to invest in hardware..
 
Last edited:
PREDATAR Control23

Just Left IBM

We are in the process now of moving to EMC Networker/Avamar due to the same reasons.
 
PREDATAR Control23

What's the pricing model?

I'm in the process of deploying Avamar, but for desktop only. Licensing for Avamar is per TB, for us the first 4TB costs ~$90k (server and disks). Afterwards, additional 1TB is ~$10k. Furthermore, Avamar is not friendly to dbase backup (even EMC sales guys admitted as much, which is nice to hear). Avamar doesn't count node or PVU, just charge $XXX per TB of storage.
 
Last edited:
PREDATAR Control23

its not PVU based say you have a blade running ESX purchase one client load as many virtuals on that it can handle same for SQL pay for one load as many virtuals you can fit on it no more counting PVU's EMC doesnt care.
 
PREDATAR Control23

Mikeatkc. Interesting software... but dont know if it can handle 6TB a daily backup for a dev env.. Our dev env is about 500 Nodes with 6TB of daily backup

our prod is 700 nodes with 11TB of daily backup.

I would add that all the backup DEV/Prod our on the same VTL and 3584. Management is aware that if we add another backup software.. they will have to invest in hardware..

But the hardware is one-time cost with no recurring annual license. Additionally, with dev environment you can spread your nodes to backup once or twice a week on different days, etc.? Furthermore, do you need to backup dev environment that often, daily? It's dev right?
 
PREDATAR Control23

We are in the process now of moving to EMC Networker/Avamar due to the same reasons.

BTW, management here is not getting rid of TSM. It's too good of a product and saved many of my colleagues' bacon many times. TSM is not going away, just used for servers/prod only.
 
PREDATAR Control23

MikeAtc: Hardware is not so one-time cost, still have maintenance to pay after 3 years if you want to.

Yeah we could backup twice a week... But it would cost the same in TSM licences... if you backup a machine you have to pay.. even if it's once a week or more..

We would have a little gain in Tapes since we keep only 7 days for DEV env and i do more BMR of dev machine than production.. it's been more than 2 years that I challenge mangement for dev machines.. They all say.. no it's important.. and when i bring the bill they say cut the cost. I give them the list I want to cut they say no.. and we start over and over..

So now i'm looking for another alternative. But when you look at other pricing ex: per TB.. You will have the same situation than PVU'S You will have 10% of your machine generating 80% of your data if you check per TB. (but more easy to control the cost i think) and per PVU, you will have 80% of your licences representing 10% of you data backup.

If you take a IIS machine, You have other ways to backup, If a VM you clone and keep the clone, you can do a NTbackup into a flat file and dump it on a servers. or install TSM or other backup software..

All have there pro's and con's Maybe having only 1 backup software that do everything might not be the best solution for us but on the other side having to much backup infrastructure is no way better.. I still work on finding the best of both world..

This is my 2 cents
 
PREDATAR Control23

This is where my environment is headed to, multiple backup solutions. The workstations are predominantly Windows 7/XP and hence will backup to Avamar, utilizing their dedup feature. The servers are staying put on TSM, prod and dev, and my servers environment is not that huge yet to concern with (yet).
 
PREDATAR Control23

I have just been dragged into the license cost debate in our environment. When I last was involved in the pricing side of this TSM was so cheap, nothing could come close. We swept NetBackup off the floor and reduced our client costs by over 80% back in 2000. Now this PVU model is going to force TSM off the floor. A backup license cost of several thousand dollars for a single server backup is not going to fly.

I am also looking at Avamar as a potential solution for at least part of the environment. For those who have used this, how do you like it? Is it more cost effective than TSM? I don't believe the marketing nonsense about 500:1 deduplication, but other than that it sounds interesting.
 
PREDATAR Control23

The above link showing the special bid based on TB made pricing very clear. If we were to deploy that model we would be paying about $1.7 million a year.

I know we aren't paying that much now so what is the best way to determine what we should be paying for licensing based on the PVU model? We have 20 TSM servers and over 1600 nodes. IBM doesn't expect us to manually gather PVU info do they?

I am aware there is a Licence Metric Tool and we attempted to deploy it but it was not without issue. At this time it only exists on 20% of our servers.
 
Last edited:
PREDATAR Control23

Yes IBM does expect you to gather PVU info. When I last complained about that to my local IBM reps they said it was planned for TSM 6.3 to have a tracking method for PVUs.

Also, by the same reps, I was told that IBM adjusts the pricing of the TB model to what you pay now, to avoid any pain points of moving to the TB model. The gotcha is IBM also, based off of your TSM primary STG pool growth history, does a forecast of where your primary STG pools will be in two years and basically creates a contract for 2 years worth of growth, aka locking in at least two years. Of course there might be other gotchas with this approach that will arise.
 
PREDATAR Control23

We've had TSM since 2006 and I've never gathered any info for a license audit. I wonder how we are being charged or how the licenses are being audited. A manual convoluted process of gather PVU probably hurts them more than helps them. I suspect most people aren’t reporting accurately.
 
PREDATAR Control23

This was the thread I started early last year when I was complaining about the process to find PVUs. http://adsm.org/forum/showthread.ph...our...-wait-is-it-dual-core-or-hyperthreading...

I believe there is another post listing this information but in my environment we have AIX, Linux and Windows. AIX is EASY to find cpu core count, Linux.... not bad just more lines of code, windows... well, at the time it was easier for me to hit every box, since a lot of our windows are VMs.

AIX command:
Code:
lsattr -El sys0 |grep ent_cap |awk {'print $2'}

(upper e lower L, it is NOT an upper i or the number 1)
Thanks to heada - listed on the link above.

Linux command (already in the format of a script):
Code:
#!/bin/bash

# find the number of real cores/cpus in a box - ignore hyperthreading 'logical' CPUs

if [ ! -r /proc/cpuinfo ]; then
        echo 'Is this Linux? Cannot find or read /proc/cpuinfo'
        exit 1
fi

# check if the physical id and core id fields are there. if not, just use
#       raw processor count as the number
num_cores=`grep 'physical id' /proc/cpuinfo | sort -u | wc -l`

if [ $num_cores -eq 0 ]; then
        # this box is either an old SMP or single-CPU box, so count the # of processors
        num_cores=`grep '^processor' /proc/cpuinfo | sort -u | wc -l`
else
        # have to factor in physical id (physical CPU) and core id (multi-core)
        # for each 'processor' in /proc/cpuinfo
        #       concatenate  physical_id and core_id  then find the unique list of these
        #       to get the # of cores/cpus
        list=(`grep -iE '(physical|core).*id' /proc/cpuinfo | cut -d: -f2 | tr -d ' '`)
        index=0
        for ent in ${list[@]}; do
                new_index=$(($index/2))
                tmp=${new_list[$new_index]}
                if [ -z '$tmp' ]; then
                        new_list[$new_index]='$ent'
                else
                        new_list[$new_index]='$tmp,$ent'
                fi
                index=$(($index+1))
        done

        num_cores=`echo ${new_list
[*]} | tr ' ' '\n' | sort -u | wc -l`

fi

echo $num_cores
Taken from dslreports I believe, originally found it on a google search

Windows command:
Code:
[URL]http://adsm.org/forum/showthread.php?17178-CPU-Core-count-one-two-three-four...-wait-is-it-dual-core-or-hyperthreading[/URL]
Thanks to THE_WIPET

These help me out, fyi.
 
Last edited:
Top