Hitachi's Hu Yoshida and IBM's Barry Whyte have had a good exchange in the blogosphere recently about the similarities and differences between IBM's SAN Volume Controller and Hitachi's USPV.
I'd like to share my perspective on this issue as an independent Wikibon observer that speaks with many customers of each product. The premise of this note is that the similarities between the two products are far greater than their differences and in these tight economic times, CIO's should have their staffs carefully look at these solutions and the business capabilities they support. Specifically, systems that offer capabilities exhibited by these products can help users improve return on assets (ROA) which we believe will become increasingly important in 2009. We see storage virtualization in general and heterogeneous storage virtualization specifically as techniques to support this business driver.
To the main point, both products support and virtualize external heterogeneous storage and are two of the few products on the market that provide this capability. To be clear, there are undoubtedly differences in the two products. Specifically, the USPV has a 'built-in' tier 1 capability allowing the use of internal storage capacity whereas the SVC exclusively supports externally attached storage.
The difference here is that SVC users often fret about placing a tier 1 device (e.g. a DMX) behind an SVC, although there are many examples of customers who do just that (successfully and normally with older DMX systems). With the USPV, the tier 1 storage is a fundamental capability built in the box and frequently we've seen Hitachi sales reps actually dissuade customers from hanging a Symmetrix as a tier 1 option off of a USPV (presumably because they wanted to sell Hitachi tier 1 storage instead). This has cost Hitachi some business in some cases and perhaps preserved margins in other. Here are two nice case studies of SVC customers (not vendor sponsored) that describe many of these dynamics:
- Reducing migration pain: Virtualized tiered storage
- Bringing Order to Chaos with Virtualized Tiered Storage
And here's a business case study on the economic considerations for users of USPV (again, not sponsored by a vendor):
Our experience, which is underscored in these case studies, suggests that SVC customers are often focused on establishing a tier 2 'default' capability (to avoid tier 1 and cut costs) whereas USPV customers are frequently killing two birds with one stone, namely supporting migration within tier 2 devices while at the same time leveraging a tier 1 infrastructure.
From a business perspective both solutions offer very capable tiered storage solutions and excellent ROA support because they allow customer to virtualize installed storage arrays from virtually any major vendor.
In our view, despite the myriad technical nuances, many of the differences in these two products come down to marketing. Frankly, Hitachi has done a poor job positioning the USPVM Diskless version, which is MUCH more like the SVC from a use case standpoint. We suspect Hitachi has avoided pushing hard with the USPVM Diskless because Hitachi doesn't want to eat into sales of its high end, higher margin USPV. We feel this has been a big mistake and a lost opportunity.
Based on Wikibon community discussions, we believe the USPVM Diskless, IBM's SVC and EMC's Invista are all aimed at the same problem, namely the need to ease migration pain and take advantage of installed assets. As well, LSI has just announced SVM 5 which HP will be reselling. In terms of where the products are at, the SVC works well and has many good references, the USPVM Diskless is a very nice package that has been handcuffed by Hitachi marketing/positioning choices and Invista has been a complete flop, suffering from technical delays and the lack of good references that have scaled the product without major problems. LSI is late to market and is just getting started.
Most amazing in all this is that the complete lack of a credible storage virtualization strategy hasn't seemed to hurt EMC one bit. 3PAR, Compellent, EqualLogic/Dell, Lefthand Networks (HP), NetApp, LSI and others have these dead simple, kick butt virtualization products and they, along with IBM and Hitachi haven't seemed to make a meaningful dent in the EMC train.
Our information suggests that more than 70% of storage users in North America are expecting budgets to be flat to down, with the down side as much as 30% in 2009. In this climate, products that are easy to deploy and manage and offer simplified provisioning would seem to be attractive. Moreover, those that offer the attachment of installed heterogeneous assets, warrant further consideration.
Action Item: CIO's should look hard at the decade long practice of putting up with cumbersome LUN management and painful migrations for SAN infrastructure. In these tight economic times these practices should be carefully scrutinized and shed where possible, in favor of virtualized storage systems. Solutions that support heterogeneous storage migrations should be considered for shops with diverse array infrastructures.
Footnotes:
