Mike,
I read your latest post about automated tiered storage, sorry but I found it of scarce quality and full of FUD.
I agree, more or less, with you about the story and theory of automated tiered storage but I disagree with the rest of the post.
First of all you are completely wrong about the concept of “enterprise storage” if Compellent (or, as you like, CML) isn’t enterprise I find very difficult to define as enterprise PLR ( or, as you like, Pillar 🙂 ).
Pillar, as probably you know, do not sport enterprise features like replicas! (Pillar uses third party software tools to be installed on the servers to do replicas). Compellent has a lot of enterprise features: unlimited snapshots, hardware and firmware upgrades without any service interruption, scalability from 5 to more than 1000 disks without changing controller model, replicas, boot from SAN, widestriping, space reclamation for windows, automatic restriping and performance scale up upon the insertion of new disks, live data migrations, and so on… can you show similar features on your arrays?
On the other hand Compellent is one of the most growing companies of the storage industry, can you say the same for yours? by the end of the month CML (or Compellent) is going to show to the finance interested audience his earnings, can you do the same?
From a technical point of view, I would like to talk about the table shown in your post:
QoS. The QoS was born to cope with bandwidth and latency limits in the tcp/ip networks. It is the ability to prioritize some kind of more valuable and latency sensitive traffic instead of other (Compellent has QoS and dedupe capabilies for remote replicas!). In a storage area networks is more difficult for me to understand why QoS should be so important: CML has data progression (automated tiered storage), fast track (data placement optimization), instant replay (continuos snapshots) all tightened in the storage Center to offer a real QoS!!! For each, or group of, LUNs you can define a storage profile. This means that access to your data is prioritized according to the tier chosen, disk position and real usage. Simple to understand, configure, manage and, of course automatic!
Schedule and signaling. So your array can move LUNs between tiers based on a schedule or when you get an alert. Intersting, but what does it mean? i.e. you can schedule your production LUNs movements back and forth between tiers in the weekend? hmm! useful! Are you sure it will help your customers? How much is it convenient to move entire LUNs: the IT history says: nothing! And what about alarms? You say you can script data movements reacting to events? who will maintain the scripts? where are they located? on hosts? with different scripting languages and/or operating systems? WOW! it will help the TCO of your customer. BTW, let me know at least 10 real implementation cases of this features so we can understand why someone would like to chose complex, risky, human dependable storage management procedures instead of industrial strength proven algorithms worldwide successfully used.
I could go further with my thoughts but i’d rather have your comments first.
Ciao,Enrico
PS: I do not dislike all of Pillar stuff, i like very much your glossy shiny brochures and Pillar’s plastic front array covers with a lot of glittering green leds, they really give a touch of enterprise look 😉