Actually, this is a very simple question: "why buy expensive storage, if it is possible to achieve the same with a much cheaper storage?"

This sounds very reasonable, but often it is just difficult to implement. Let´s have a closer look on a typical top level target system for storage planning:

  1. Reliability / safety
    The storage must be reliable – data loss is not allowed to happen at any time.

  2. Reliability / performance efficiency
    At all times the storage system needs to have enough performance reserves. Bottlenecks which are generated by too weakly dimensioned storage systems should be excluded at any time.

  3. Expandability
    The growth of the storage system is a fact and demands more capacity and a higher performance from year to year. So the system´s expandability should be kept easy.

  4. Cost / follow-up cost
    These goals have to be achieved with the smallest cost-effort possible. During the planning period, follow-up cost for maintenance, space requirements and energy consumption have to be considered.

The parameters reliability / safety can nowadays be achieved on different cost levels with various storage classes. Low range storage, which previously was perceived to have the same technical unreliability like SATA, has been improved in its quality.

In contrast however, the parameters of reliability / performance are usually still diametrically opposed to the cost. Especially if one considers the unknown, a general unknown parameter which obviously has an important role in every storage planning - the uncertainty of the performance planning.

The justified concern, to run into a performance bottleneck today or tomorrow, misleads many responsible employees to oversize their storage environments consciously. A supporting component is certainly also the currently much exaggerated usage of SSD storage, which is touted as a panacea to cure all performance problems in the world. Although SSD is getting cheaper, it still costs a multiple of disk storage.

In general, information about the current load and empirical data which might arise from observations is lacking. Where this reliable information is lacking an estimate is made associated in large parts with security thinking. In this way, spoken metaphorically, expensive storage palaces with golden doorknobs are built where actually a solid planned multi-family house would have been sufficient.


Two customer examples from big to small

This 2.5 PB system consists of 6 SVC clusters. The IO density analysis shows that almost all volumes in the cluster don´t benefit from the obtained performance opportunities which are provided by the used storage systems. Since almost all storage pools are occupied by more than 80% they have to be extended in the near future. The clear recommendation here is to invest into cheaper storage in the next expansion step.

Fig. 1: in this graphical representation of the storage system an excellent overview about all volumes and how they are supported by the used technology is given. Blue areas indicate that here the technology and the associated financial resources are significantly oversized. For this reason blue areas are preferred candidates for cheaper storage classes. In this example the 2.5PB can be recognized at a glance and it is possible to estimate that at least 60% to 70% or about 1.7PB can be stored in more favorable storage in the future.

The picture gets even clearer, if an analysis is subsequently executed to mark only volumes having a performance characteristic which can be provided by a low range storage. Here you can see exactly where the journey should go for this customer. All marked capacities are candidates for cheaper storage classes.

Fig. 2: one more time the analyzed 2.5PB environment to find out which storage areas are suitable for dedicated low range storage. With the knowledge that the majority of the used storage types come from the 15k class, the potential savings in this environment can be easily calculated.

The same happens even in smaller environments with 32TB instead of 2.5PB. In the following the IO density analysis of a smaller Storwize is displayed. This image was made with BVQ version 3.0, in which the storage level of 20% to 80% was parted into smaller steps. Again it gets obvious that surprisingly just a few IOs are needed in the storage backend. Reasons for this are on the one hand lower requirements as expected, or on the other hand the very positive impact of the cache.

Fig. 3: the color-coding of the IO density analysis (heatmap). The particularly high quality of the BVQ heatmap analysis is the fact that with the consideration of IOs, their current cache efficiency, RW conditions and other factors the cache-load is calculated. Therefore the results describe a backend IO and thus nothing else than the volume´s load on the backend storage.

Fig. 4: a relatively small environment with exactly the same results. The only difference is that the dimension of the saving potential increases with the capacity. Again, it can be assumed that 70% of the data are stored too expensively. But even here an analysis is worthwhile because the cost and follow-up cost for maintenance and operation will exceed by far the cost for analysis including software cost.

The sharpness was reduced in this image to protect the customer-specific data. But it is still obvious that far more than 70% of all capacities utilize the existing performance potential with just less than 50%.

The BVQ heatmap is the key to success

The BVQ heatmap is our tool to quickly establish an appropriate overview about the storage´s performance utilization. The heatmap can be applied to all levels of the system. A consideration of the backend storage arrays, the single managed disks or even higher grouped objects like the level of applications or entire data centers (BVQ Accounting Package) is possible.

Fig. 5: a classic example of the heatmap analysis which enabled extremely high cost savings for one of our customers. Via the analysis it was possible to detect very quickly that an extension of the very high quality enterprise systems was not necessary. More than 60% of the systems´ capacities could be released. The investment was completely turned into *nearline. Comparable to a pipe, 60% of the* high-end storage volumes were moved to the midrange or the new nearline class depending on the requirement.

The specificity of the heatmap is the fact that it may be calculated on the basis of different time periods. It displays a comparison between the current load and theoretically achievable values. In the new third version of BVQ the treemap can be set together with the heatmap to any period in the past. This makes it infinitely useful when it is used during a bottleneck analysis. It makes it possible to determine at a glance that not the existing infrastructure but another effect is responsible for a bottleneck.

What are the financial implications of such analyses?

If such scenarios are calculated with a three-set the conclusion can quickly be made that this form of analysis is worthwhile and it would be a careless neglection, if it is not performed.
The attempt to express the benefits in financial dimensions is very difficult and the result can only be expressed as a conservative estimation.

  1. storage prices are extremely volatile and may differ depending on the customer (regardless from manufacturer)
  2. For a first cost estimation a difference between near-line and high-end storage of properly € 700 or $ 900 is assumed.
  3. Please use your own cost differences in order to calculate your potential savings!

Situation – major customer wants to scale from 2.5PB to 3PB

  • The analysis shows that there is no need any more to procure high-performance storage because many areas can be moved to low-cost storage
  • Now 100% highly capacitive will be procured instead of the usual procurement of 70% high-performance and 30% highly capacitive
  • This allows potential replacement cost savings for the 70% high-performance of (with a assumption of 700 € price difference):
    • 500TB * 70% * € 700 results in € 245.000
    • 500TB * 70% * $ 900 results in $ 315.000


  • Further savings in the procurement
    • Discs have a higher capacity, so fewer discs and less enclosures are needed
    • Future savings of current expenses
    • Less need for space because of reduced enclosures
    • Reduced energy cost
    • Reduced maintenance cost


And the savings are not even finished yet!

As pointed out not only reduced procurement cost arise but also reduced follow-up cost.
Because of the possible extension from 3PB to 3.5PB in the following year, the savings effect is even accelerated over the years.

How much can be saved in a much smaller environment?

Simplified the savings are proportional to the size of the storage environment. If the customer mentioned above can save € 245000 in the first step, then it should be possible to save € 100000 to the same extent in an environment with 1PB. In the 30TB environment from the second example, in this way only € 3,000 will arise calculated with our formula (the first year). However, it must be considered that € 700 price difference cannot be expected here, because smaller customers generally have higher storage prices. Not to forget: also the analytical instruments are much cheaper here than for large environments.

And what is the key to gain all this?

The key elements are the transparency and the analytical methods which are enabled by BVQ. BVQ is the only product which is able to represent these relationships for SVC and Storwize with this high clarity and speed.
The saving-effects achieved by a BVQ solution are not just limited to reduced procurement and operation cost.
More saving effects like:

  1. Reduce operational risks
  2. Avoid performance bottlenecks or quickly control them
  3. Proactive problems avoidance
  4. Creation of added-value for the storage


BVQ web pages

BVQ in the Internet

If you are interested in further information or you want to become a partner to distribute BVQ to your customers,
please contact us at:

BVQ is a product of the SVA System Vertrieb Alexander GmbH