live volume « Cinetica

Articoli con tag ‘live volume’
CINETICABLOG

Vplex, perchè complicarsi la vita?

Lunedì scorso EMC ha presentato il Vplex durante l’EMCWorld.

A prima vista, dopo la presentazione e i mirabolanti effetti speciali, ti sembra veramente che questo oggetto cambierà il modo con cui si è abituati a vedere lo storage e che darà il via alle vere private cloud, poi però sono usciti la documentazione ufficiale e un pò di post ed è facile rendersi conto che è solo uno specchietto per le allodole!

Il Vplex è un appliance che deve essere inserito “on top” ad una SAN permettendo quindi la sua virtualizzazione. Il vantaggio più grosso (forse l’unico?) che ha è quello di avere una cache particolarmente efficiente (derivata da una acquisizione di un paio di anni fa) capace di distribuire le scritture anche su due SAN in remoto (fino a distanze 100Km con meno di 5ms di latenza) garantendo quindi la possibilità di avere una LUN che è contemporaneamente accessibile sia localmente che remotamente.

WOW! infatti la prima demo che hanno fato vedere è quella più impressionante, Vmotion di centinaia di server in pochi minuti fra due datacenter diversi.

Non storage Vmotion, che avrebbe mosso tonnellate di dati con tempi biblici, ma una semplice Vmotion! Facile, veloce, indolore.

Ad avere il budget e l’infrastruttura adeguata significa migliorare la qualità del servizio in modo sensibile eliminando ogni tipo di fermo ma purtroppo non è così: Il Vplex non ha le funzionalità di base che uno si aspetterebbe da un oggetto del genere e quindi è praticamente inutilizabile.

Basta pensare a tutte quelle funzionalità che troviamo normalmente nei virtualizatori (SVC di IBM, USP di HDS o Vseries di NetApp): Thin provisioning, tiered storage management, snapshot e così via.

Insomma, più che una soluzione sembra un esperimento di laboratorio quindi mi chiedo: perchè complicarsi la vita e aumentare il TCO di tutta la tua infrastruttura quando puoi avere le stesse funzionalità direttamente dal tuo Storage?

Ovvimente mi riferisco a Compellent e al suo live volume.

Fra l’altro vplex costa di più che migrare da storage EMC a Storage Compellent.

Tag:Compellent, EMC, live volume, Storage, TCO, Vplex
Pubblicato il 13 maggio 2010 nella categoria Compellent, Storage | 2 Commenti »

Just another V-(per)Plex(ed) post

Well, EMC announced V-Plex during their EMCWorld.

IMHO, It’s the storage federation seen from the EMC point of view: adding a lot of complexity with little value to screw the customers as much as possible. Seems to me the same old story of the last years: “this is federation V1 (like FAST V1) it isn’t nothing special but it is branded from EMC so it’s cool”,

In the latest hours I read a lot about V-Plex (here, here, here, here and here are only few examples) but my first impression is that EMC hasn’t achieved an outstanding result, they only reinvented hot water (i.e.: SVC from IBM, V Series from NetApp and USP-V from HDS) adding some interesting capabilities like “VMs Teleporting” that you can also find embedded in Compellent’s Storage Center with its Live Volume (who’s been at this year’s C-Drive knows what I’m saying.).

Why V-Plex is wrong.

V-Plex is another layer of Very high Complexity (probably the name comes from here), because you need to place one (or more than one) new appliance between your SAN and your hosts to get this virtualization layer. This approach has some advantages and you can find all the related white papers and documentation from many vendors, especially if you want to reuse old and non virtualized stuff but, at the same time, a lot of disadvantages:

You need to acquire new hardware, support and services;
You need to maintain both old and new hardware (support, services, etc.)
You need to train people;
You need to hire consultants and/or EMC professional services ($$$$$);
You need to avoid constraints and limitations in feature interoperability!
You need a lot of human work: migration to a V-Plex based infrastructure is not automatic!
You need more space, power, cooling in your datacenters;
and so on…

The final picture is quite interesting but the cost to paint it is very high! Why don’t integrate these features directly into all your families of existing arrays? Compellent does!

TCO of a “v-plexed” SAN grows a lot without adding much if compared to other solutions.

Some questions come to my mind (but I admit that I need to further investigate before a final comment): My thoughts are about the fact that V-Plex is a virtualizer with a pretty nice cache (indeed, the caching technology doesn’t come from EMC labs but from an acquisition, namely Yotta Yotta) but you need to understand very well some basics before deploy this new object on top of your existing SANs, the questions are too many for this post but there are terms that never appears in the white papers: thin provisioning, snapshots capabilities, different raid levels (i saw only 0 and 1), automated tiered storage (FAST in EMC parlance), all the replica capabilities, etc…

In other words Virtualizers are very useful to harmonize your SAN infrastructure but if it is only a cache you will add complexity to complexity cutting down the advantages and adding costs!

V-Plex isn’t Federetion (or private cloud enabler), it’s more a Virtualization Appliance!

In the last days we’ve seen interesting posts from the blogging community about storage federation and one of the first point of discussion was on the fact that an appliance on top of storage doesn’t mean federation. If you imagine federation as a bunch of arrays working together as a whole system managed by a single pane of glass you are not thinking to V-Plex.

V-Plex is a second generation virtualizer/federator experiment generation, after the failing InVista, and it will be better than its predecessor for sure, but I don’t think it is a good solution for a couple of reasons:

TCO and TCA are very high: The cost of a base appliance starts at 77K$ and I’m very curious to know the real expense for the total solution delivered on two data centers, training costs, maintenance, professional services, migrations to add on top of your existing infrastructure.
Real Value: As of today I haven’t perceived the real value of the solution! Non disruptive LUN movements? It can be done for a fraction of that price with a feature of the Compellent Storage Center called Live Volume.

The viable alternative

V-Plex is an high-end solution with a lot of maturity problems and high costs, but EMC is targeting the right point: there are some needs to be addressed in modern environments. EMC can’t do this with its actual storage array proposition so they are trying to sell you an appliance to do that, Compellent has smarter arrays and they can do what you need just by enabling a software feature:

Non disruptive Lun migrations between arrays;
Virtual machines migrations (not only VMware but also Citrix and Hyper-V);
Disaster avoidance;
On-demand load balancing;
Zero downtime site maintenance (i.e. power outages) and server maintenance;
Enablers to build a cloud computing architecture

V-Plex target is higher? I don’t know but who cares?

Live volume solves 99% of your pain points, it is easy to use, fully integrated, without hidden costs and at a small fraction of the cost!

It’s a good solution also if you are the owner of CX arrays: swapping Clariion CX arrays for Compellent Storage Center costs less than the acquisition and implementation of V-Plex.

We will do other posts, in the near future on live volume, it’s an amazing feature and I think people need to know the difference between complex and smart storage management!

You will find other informations o Compellent web site very soon! Here you can find the first infos from Compellent.

So why buy a V-Plex ? when you can have Compellent ?

Tag:Compellent, EMC, live volume, TCO, Vplex
Pubblicato il 13 maggio 2010 nella categoria Compellent, Senza categoria, Storage, Virtualizzazione | 3 Commenti »

Can you do the same?

Mike,

You don’t publish on your blog comments from people who do not think your same way, ok, it’s not good but hey, it’s your blog. I don’t understand your hate (or simply envy?) against Compellent, I don’t know if you have an installed base here in Italy, (I found pillar just once and I won!), so i can’t verify anything first hand about Pillar.

Well, you are gearing up with FUD! it’s funny for me, I will answer point by point to your misleading post to try to put some more reality into this “discussion”:

1) I have had a lot of Compellents installed in enterprise class environments with Solaris, Linux, Windows, VMware, and one Oracle VM! I saw disks crashing without affecting the customer, error messages in the logs of the hosts nor timeouts of paths or LUNs. If a path doesn’t respond for more than something like 30 seconds a timeout will accour, I never saw this behavior! And i don’t remember something similar for LUNs.

FUD 1: where did you take your data from? Moreover I saw better rebuilding times than sported by your graphs! please let me know more on the compared systems, from Pillar and Compellent. Are they really comparable, and well done configured with similar workloads, used capacity and type of disks?

The Real World 1: I saw a complete rebuild of 1TB SATA disk (out of 45 rotating disks) full at more than 80% in less than one day, without evident performance impact to the hosts! Are you sure you have done the right work with the Compellent tested? I have the logs of the Enterpise Manager, the Compellent’s management tool can produce a lot of logs, reports, and graphs: i can show them in public! can you do the same?

2) We have Data Progression installations all around Italy with different kinds of workload, operating systems, and type of disks installed. Never seen or heard about what you are saying. Customers are all satisfied and, according to indipendent analysts surveys, Compellent can show a 98% (and growing) level of satisfaction: can you do the same?

FUD 2: It’s very difficult to tell a potential customer that you don’t have the features he is asking for! But it ‘s foolish to say “that feature doesn’t works properly!, i heard about outages somewhere in the outer world!!!” Hey, i have heard about outages of HDS USPs in the last months (very well advertised and in the inner world): but, can you tell USP is it one of the worst storage systems of the world? and more, When you will be able to release your automated tiered storage v. 1.0 Compellent will release the version N.0th version! BTW, wait just a while before speaking about strange and not proven issues on a feature you haven’t already released…. are you sure you will be flawless?

Real World 2: Compellent customers use data progression (automated tiered storage) since years, they are very satisfied and they continue to buy Compellent (take one more look at the surveys, please!). Can you do the same?

3) WOW! Pillar can upgrade firmwares (seems only minor releases) without outages? Compllent does it very well since years, where’s the piece of news?

FUD 3: I personally saw firmware upgrades without service disruption, change of number and quality of ports, recabling, hot restriping, and other activities without any inconvenient. can you say the same?

Real world 3: hot firmware updates (not only point releases but major releases) are an ordinary job! Compellent can migrate from an old controller to a new one without service interruption! can you do the same? Compellent can live temporarily with different controllers generation in the same system, can you do the same? Compellent can migrate LUNs between systems without service interruption (live volume) can you do the same? Compellent has ONE system capable to scale from 5 to 1000 disks without needs of forklifts upgrades, can you do the same?

4) A statement like this one does not make sense for a Compellent, so i have nothing to answer. Do you have problems with your systems in that way?

Real World 4: Compellent is well designed, I never experienced what you are saying, probably you are talking about a bad configured one.

5/6) hard to think about a technician with a solder in a Data center, where did you see one? Never saw this about Compellent but ok, FUD is FUD!

FUD 5/6: Compellent is a Standard hardware vendor (more standard than Pillar), if you are guessing about the fact that the Controllers are standard x86 servers. Yes, they are! and this is one big advantage!!!!

Real World 5/6: The big advantage is that standard hardware (developed by tier 1 vendors like Qlogic, Seagate, Intel) is more realiable, bugs free, cheap, already tested at 100%, developed by multibillion dollars companies with huge R&D budgets! Compellent is doing only integration. Can you say the same for your hardware? Can we trust in you with your poor installed base? or in your small R&D department? (i would like to remember to who is reading that Compellent is hiring while Pillar is firing! ). Standard hardware, I wrote a lot about this in the recent past, allows Compellent to spend more in development of software features (asked by the customers), Compellent uses standard hardware so they don’t need factories but use the multibillions dollars companies ones to assure quality and response times, can you do the same?

BTW, you have not answered to my latest post about features from Compellent that we can’t see in Pillar’s arrays and about your financials.

ciao,

Enrico

PS i’m a Sun Reseller since 1996 and i never mentioned or heard someone referring Sun Microsystems as JAVA!

Tag:Compellent, Data Progression, live volume, Pillar, Storage, Storage Center
Pubblicato il 26 ottobre 2009 nella categoria Compellent, Storage | 4 Commenti »

Vplex, perchè complicarsi la vita?

Related Posts:

Just another V-(per)Plex(ed) post

Related Posts:

Can you do the same?

Related Posts:

Ricerca

Subscription Options:

Twitter

Post recenti

Archivio

Links

Related Posts: