jeudi 20 décembre 2007

The other day a vendor came and told me he wanted to help me solve my storage problems:

Vendor: we want to help you solve any storage problem you have
Customer: how's that?
Vendor: because we have a solution approach, we can help you solve your problems
Customer: right, you bring solutions to my problems! so here's one of my problem: How can I honour storage requests coming from various groups
Vendor: easy, buy our SRM tool!

The term solution can be ambiguous, as it can either mean
1. the answer to a problem:
the answer to my storage provisioning problem would be a SRM tool
2. the process of solving a problem
the key word here is process: what "things" can I put in place in order to solve, on a long term basis, my particular problem

As a matter of fact, as a customer, I was after the 2nd definition of "solution", when this vendor offered me the 1st one.
What I was looking for, was a process to solve my provisioning problem. A good process survives any technology.

In other words, the answer to my problem is a building block, part of an IT architecture

Mister Vendor, listen to me, an IT Solution is either
* building block, which is part of a framework called: IT Architecture.
* Service defined as part of a service catalog

If you own the architecture, you own the building blocks and all the components that go into them. Otherwise, just sell the components that make my building block, but don't expect to be my trusted advisor.

For an introduction to IT Architecture check this out

Next blog, I will post a table of content of an Storage Architecture document. keep coming!

mercredi 12 décembre 2007

The time is coming

CIFS server for Opensolaris is now reality: first release is available here
anytime soon, it will be integrated into Solaris its self and : bye bye Netapp!
Company starts to undestands limits of Netapp, and this is a fact I am not allowed to prove
but we are near the time where, from a simple DVD, one will be able to install, on a commodity server (as well as a high end server) a fully functional NAS appliance, with all advanced features such as snapshots, ndmp backup, CIFS/NFS (without addditional cost), FC (as Lun provider), and even archiving features such as SAMFS

samedi 1 décembre 2007

Parmigiano for my Spaghettis

I invited 2 colleagues of mine at a pasta restaurant called "Netapp" to try their speciality "Spaghettis" (see previous post). I has happily eating the superb pasta with my fork (no spoon!) and suddenly I felt I needed to add Parmigiano ... I asked the waiter to bring me some when he announced it would cost me...$47.000 to add "cheese from Parma" on those spaghettis !!

  • Me: What?? you must be joking! $47k to use a stupid cheese??
  • Waiter: Mi dispiace, parmigiano for 3 plates cost $47.000
  • Me: no thanks, next time I need parmigiano, I'll choose another pasta restaurant..crook!

Frankly, do you think Netapp is really entitled to ask $47.000 to use NFS ??? is this fair when NFS has been around for as long as unix exist.

Basically I'll wait a couple of month to have more recent boxes doing NFS without additional licenses (may be fishworks or Nexantastor)

bon apetit!

mardi 20 novembre 2007

Old spaghettis with a lovely sauce

ok, let's face it, Netapp boxes are ugly: spaghettis, daisy chain connections, 32bits snmp..yuk

but Ontap is really great, no wonder why some companies want to copy it using commodity hardware/OS (check nexenta or opensolaris with Fishworks project)..OnTap is a success and by far.
I really appreciate how easy it is to configure (once you managed to untangle cabling):
  • create a trunk (called vif) of four Gigabit Ethernet links is very easy and it works
  • clustering is straitghforward...and it works
  • GUI interface allows to do 99% of the tasks (ok the interface is ajax or anything dynamic)
  • snapshot, ndmp backup, snapmirror works first time
  • netapp simulator is great as it helps testing configurations

so, although spaghettis are old..the sauce is really good!

mercredi 14 novembre 2007

old spaghettis

did you know that Ontap doesspeak snmpv1 only...i.e in 32bits. so snmp walks are nightmare when dealing with large sizes (above 2GB). Ok it simulates 64bits by sending 2x32bits counters that one has to re-assemble..of course nobody bothers, it's too complicated to modify a monitoring software to do that
also, the spaghettis cabling IS a nightmare: even Netapp engineer had to think hard before we could see all the drives.
last but not least, you cannot swap tray orders: some of them contain the OS, and they must be placed correctly in the scsi chain if you want to boot successfully (especially in a cluster)
these spaghettis are really old
ideally, tray connection should be done as a SAN fabrric topology, not a chain. that eases management and storage expansion (basically ou would not need to break the chain to add a storage tray)

lundi 12 novembre 2007

Cold Pizza

We just received delivery of our new pizzas: made by the pizza market leader EMC.
our pizzas are made of 8 nodes, with linux as basic toppings, and with some special software on it, which nobody understands, and nobody can tell me where it comes from. this pizza has a name: Centera

Does anyone know where Centera comes from? was it built by a clever team of EMC engineers? is it the result of an acquisition?

Well I'll tell you: in fact Centera is the product of a $50Million acquisition of a belgium company called "FilePool NV" the way the name FilePool appears at boot time of a Centera node..acquisition happened in 2001

Have a look at this old document (1999) explaining how FilePool works

Here again an old technology, cleverly marketed, which I believe has made his time. and is still beeing sold

Now during install, one of the node did not work (could not join the centera cluster). So we called EMC for a replacement , here's and excerpt of the conversation:

-EMC: connect the box to the phone line and will troubleshoot
-The Bank: no way, the centera will not be connected to a phone line
-EMC: this is not how it works sir, support must be done remotely over the phone
-The Bank: This box will hold client information, impossible to connect it to the outside world and accessed by an external company. It's against the law
-EMC: well then, troubleshooting can take up to a month to be resolved

imagine that, we spent $200k for 10TB (raw) of data and all we get is several weeks delay for a node replacement....which will not even been done by EMC, but a partner (Unisys).

we get much better service at the pizzeria: if the pizza is cold, we get a replacement within minutes!

mercredi 7 novembre 2007

Do you like Spaghettis?

well if you like bolognaise, try this recipe:

take a NetApp FAS3020c (cluster), with 4 disk trays and cable everything with fibre-channel cables ...power on, don't forget to set the terminator and the tray IDs manually (yep! like in the old days) voilà!

Here's a diagram for appetizer

you end up with more than 16 cables to interconnect the lot and 10 ethernet cables..all this for a mere 4TB of data!

gosh these machines are so old-fashion, I thought the time where we had so many cables+trayIDs where revolved for good..but no, NetApp is really getting old

bon appetit

dimanche 28 octobre 2007

I am not hungry for food!!

I did my "Storage Vision" presentation to a larger audience last week: all the IT organisation was there (70 people showed up, out of 150...basically storage interests people! ). The subject was about my storage strategy and implementation planning ..that was fun
Considering I work in an English bank, and that Rugby world cup final was last week-end (England vs South Africa), I used the theme of rugby as a background analogy:
  • Data is to Storage, what "oval ball" is to rugby: an entity that must circulate among the players, must be protected, and the goal is make value (score) out of this ball.
  • Rugby is a playfield, rules, players and referees. Storage Architecture is: infrastructure, process, procedures, building blocks, governance
  • Rugby game is nothing without a ball, storage is nothing without data
  • so forth...
Sport analogies are easily used...

The long term strategy I presented was: move to a "service oriented storage architecture".

So basically:
- put in place necessary tools & processes
- put in place a storage service catalog
- invest in tools to streamline storage demands
- tools to report accurate storage usage (gosh how difficult it is in a large environment
- manage risk
- manage change
- manage performance

Managers and stakeholders like the idea, but not all my fellow colleagues: I sense some of them are already taking their distance with I presenting a threat somehow??

I proposed to invest in a product called SANscreen from Onaro (if you don't know this product, go to their website right now, or buy the company ASAP!!). Product costs 400kUSD (approx. list price) to manage 1500 SAN switch ports, do the reporting and do the storage reservation. It is not a monitoring tool, but a SAN Architect tool, and thus it is not replacing what we already have (HiCommand)

Price did not shock managers, but rather my colleagues:
"what!! you just arrived and you want to spend all this money in a monitoring tool???"

...when you know that, first these guys develop their own tools (based on opensource, costing only extra-time to the company), and second they fight like hell to obtain a new workstation.. you understand their reaction. Of course, an International Bank refuses proliferation of in-house open-source, not supported solutions..this point was made in a recent internal communication, furthermore it's written "black-on-white" (french expression) in the "IT Finance and Planning" document.

Talking of opensource, I did not mention IBM's project: Aperi
However, to maintain a "good and saine" relationship with Onaro, I will start a long-term project about evaluating Aperi as a SAN management alternative. Opensource is a good tool to maintain competition!

Also, I've been given the task to define the new Backup/Restore infrastructure, and do the migration. So we are going for a couple of SL8500, 20xT10k tape drives, a couple of VTL-V..all this to be in production quickly...budget 1.5m USD
Still don't know if I'll buy PS yet..who would do the job anyway?

At the same time, I am doing the planning and the build specifications for a couple of Netapp FAS3020c (sold by an ex-Sun) and a Centera stupid-box (also sold by an ex-Sun)

Basically, at the moment I am helping Sun to sell these boxes other words, Sun is making more money by having me working for a customer..isn't that ironic...don't you think?

One of the advantage of being a customer is to be in touch with all these vendors: IBM, HDS, SUN, NETAPP, EMC..drawback, they all want to get me fat. Listen vendor: "I am not hungry for food!!"

mercredi 10 octobre 2007

storage architecture vision

yesterday I presented my storage architecture vision
for the bank. It was well perceived, and here's the recipe to write such a document:
- start by finding the right documents, and who wrote them (owners). Initial document inputs are the global strategy of the company, IT principles, and all the architecture documents you would find
- Write a first part, relating to how these inputs apply to storage
- List all the storage building blocks and develop how you would use them
- Define a set of rules for these building blocks
- Define the services you provide to your "customers" (in this case, customers=internal groups)
- draw a nice big diagram
- Present everything in front of the stakeholders

et voilà

jeudi 4 octobre 2007

Gimme a price!

Principle 1:
A vendor works with several cutomers
A customer works with several vendors

A customer has access to (valuable) information, a vendor would like to have: competitive information
A vendor has information a customer would love to have (expertise)

A vendor such as Sun/IBM... spend a load of money to get competitive price info. Here's what happened to me:my boss gives me a budget of USD250K to find an SRM product. Well, I have the budget, I know what product I want; off course, the budget is not to be given easily, I must first compare prices among potential vendor. I'm lucky, the product that fits my needs (AppIQ) is available from HDS, HP and Sun, I just need to ask a quote, for the same configuration:
- HP gives me USD350K
- Sun gives me USD1M (yes..that's not a typo)
-HDS, well they are not keen on giving me a price, coz the product is "too complex", I should try to get something else, made by HDS...a crap USP centric soft basically.

Problem description:
How to explain the price difference between HP and SUN? I gave same info to both (1000 FC ports, 500 servers dual attached)...I havent got a clue, but someone knows something I don't..

so basically, a customer has info (and money) vendors would love to get....but vendors have info, a customer would love to get also. If the interface vendor-customer is not right (trusted advisor) then we get these kind of excess

that's all folks