SSD and HDD; A Symbiotic Integration of Technologies.

Read More >>

In a recent white paper “Flash and HDD – Symbiosis or Survival of the Fittest” Tom Coughlin presented an interesting case that Flash Storage and HDD technologies are complementary and not competitive, at least from a consumer electronics perspective. The argument he presented suggests that the growth of both Flash and HDD’s are interdependent with each supporting and promoting the growth of the other. However, does this hypothesis hold true in the enterprise class data storage environment?

Over the past few years solid state storage (SSD) has regained its once fashionable standing as a legitimate data storage option encouraging the somewhat familiar refrain by many industry pundits that HDD’s were dead and within the next 5yrs or so will be replaced by solid state storage. Excuse my bluntness but hogwash. Yes SSD’s will become prevalent and yes the technology will command a significant position in the storage hierarchy, “Tier 0” if you will but the assertion that it will eradicate HDD’s is reminiscent of the many online bigots who 20 years ago declared that tape was a member of the technology walking dead. Last time I looked however, the tape market was still a viable storage option. So let me repeat, hogwash!

As a proof point check out a recent white paper that appeared in the Intel Technical Journal Volume 13, Issue 1; “Solid State Drive Applications in Storage and Embedded Systems”. This paper is a collaboration between Dane Nelson of Intel and Sam Siewert of Atrato and presents an interesting approach to intelligent SSD/HDD integration that results in an autonomic application aware architecture that is characterized by exploiting the performance of SSD with the cost balance of HDD’s, orchestrated by the block level, activity and access profile, of data.

Some of the key features of the architecture discussed in the paper include:

1. A hybrid solution that can scale to petabytes.
2. Ability to profile I/O access to petabytes of data using megabytes of RAM.
3. Ability to create an SSD VLUN along with a traditional HDD LUN with the same RAID features so that file-level tiers can be managed by applications.
4. Ability to create hybrid VLUNs that include HDD capacity and SSD cache with intelligent block management to move most frequent accessed blocks between the tiers.
5. Ability to create hybrid VLUNs that are composed of HDD capacity and are allocated SLC SSD ingest FIFO capacity to accelerate writes that are not well-formed and/or are not asynchronously and concurrently initiated.
6. Ability to create hybrid VLUNs that are composed of HDD capacity and allocated RAM egress FIFO capacity so that the black-end can burst sequential data for lower latency sequential read-out.

This I suggest is an example of the intelligent and thoughtful integration of storage technologies that extends Tom’s original consumer electronics based hypothesis to enterprise class storage, that Flash and HDD technology is symbiotic and not competitive.

60 Storage Facts, Figures and Estimates

Read More >>

Many thanks to Fred Moore of Horison Information Strategies who compiled these facts from a wide variety of IT industry sources.

1. Average annual digital storage demand rate (primary occurrence of data, all platforms)
   i. 35-40% for all data (2010-2011)
   ii. 20-25 for production data
   iii. >60% for fixed content, compliance and archive
2. Amount of digital data stored on Unix, Windows and Linux systems WW
   >90%
3. Average disk allocation levels for z/OS (eSeries mainframes using DFSMS suite)
   60-80+%
4. Average disk allocation levels for iSeries (AS/400 servers)
   60-75%
5. Average disk allocation levels for Unix/Linux systems
   30-45%
6. Average magnetic disk allocation levels for Windows systems
   25-40%
7. Average annual disk drive areal density increase
   35-50%
8. Average annual disk drive performance improvement (seek, latency and data rate)
   <4%>
9. Increase in disk drive capacity per actuator since the first disk drive in 1956
   400,000x (5MB to 2000 GB or 2 TB)
10. Increase in native tape cartridge capacity since the first tape cartridge in 1984
    5,000x (200MB to 1,000 GB or 2.0TB compressed @ 2x)
11. Recommended data center power consumption level
    100 - 500 watts per square foot
12. Power usage breakdown in typical data center
    Chiller – 33%, IT gear – 30% (servers = 35%, storage = 30%), UPS – 18%, AC – 9%
13. Average cost to build a Tier 3 data center
    ~$500 per square foot
14. Electricity consumed by hi-density blade servers
    >7kW/rack and > 30kW/enclosure
15. Annual average rate increase in US for electricity
    15 - 40% (depending on geography)
16. Who gets the IT energy bill?
    Facilities team – 56%, IT team only 3%
17. Average tape cartridge utilization levels for integrated virtual tape systems
    60-80%
18. Typical range of non-mainframe disk data managed per administrator
    500GB – 28TB
19. Typical amount of disk data managed per administrator (z/OS, mainframe)
    >75 TB
20. Estimated range of automated tape data managed per administrator (all platforms)
    40TB to >1EB (varies widely based on library size)
21. Percentage of storage staff (non-mainframe) time spent on storage provisioning
    22%
22. Percentage of storage staff (non-mainframe) time spent on storage migrations
    13%
23. Percentage of storage staff (non-mainframe) time spent on backup administration
    11%
24. Annual growth rate of unwanted e-mail message traffic
    >350%
25. WW number of mainframe machines installed (2008)
    8,000 – 9,000
26. WW number of mainframe sites
    ~4,500
27. Maximum possible distance from primary data center for synchronous replication
    50 miles
28. Number of new mainframe customers since 2000
    500
29. Percentage of customers retaining e-mail archives over 7 years
    9%
30. Percentage of all e-mail traffic that is unwanted material
    ~90%
31. Percentage of attacks form e-mail virus
    68%
32. Percentage of all companies that don’t expect to move data to the cloud
    61% (July 13 09 CW survey – p27)
33. Where does the CIO report?
    41% to CEO/pres, 23% to CFO, 16% to COO
34. Percent of companies citing employees as the most likely source of hacking
    77%
35. Percentage of US adults with more than 200GB of storage capacity
    10% (approximately 30 million)
36. Percentage of digital data stored on removable media (primarily magnetic tape)
    ~75%
37. Percentage of digital data stored on mobile (portable/personal) technologies
    50-60%
38. Number of new (1st round) data storage companies funded in 2000
    92
39. Number of new (1st round) data storage companies funded in 2007
    5
40. Number of storage acquisitions in 2008
    67
41. umber of new small businesses created in the US in 2005
    550,000
42. Average revenues of InformationWeek 500 companies in 2008
    $11.03 B (was $9.0B in 2004 and $12.4B in 2001)
43. IT as % of revenues in 2001 Fortune 1000
    3.88%
44. IT as % of revenues in 2008 Fortune 1000
    2.80% (declining as a % of revenue)
45. Percentage of IT organizations managing more than 10TB of disk data
    39%
46. Average percentage of IT budget in the US spent on IT salaries and benefits
    32%
47. Average percentage of IT budget in the US spent on compliance
    5%
48. Average percentage of IT budget in the US spent on hardware
    16.3 % (projected to be 13.9% in 2011)
49. Average percentage of IT budget in the US spent on services
    23.3% (projected to be 29.7% in 2011)
50. Average percentage of IT budget in the US spent on personnel
    19.8% (projected to be 12.7% in 2011 - only if health-care costs are not covered)
51. Average percentage of IT budget in the US spent on support and maintenance
    4.3% (projected to be 4.8% in 2011 - includes energy)
52. Number of mid-market firms in the US in 2005 (100-999 employees)
    93,876
53. Percentage of all IT jobs in businesses with fewer than 99 employees
    72%
54. Percentage of WW IT workers considered mobile in 2009
    30%
55. Projected size of India’s IT services industry in 2010
    $60 B
56. Percentage of businesses who perform a regular interval DR testing plan
    52%
57. Percentage of businesses who test DR plan once per year
    26%
58. Percentage of businesses who only test mission-critical apps for DR
    47%
59. Percentage of businesses who test all apps for DR
    22%
60. Percentage of businesses who take backup tapes offsite daily, weekly, and monthly?
    Daily – 56%. Weekly -32%. Monthly - 4%.
    

Five Reasons why SAN and NAS may not meet the needs of SMB's.

Read More >>

In an earlier post I talked about the seven key attributes of SMB storage but what I did not talk about was the storage options available to the SMB.

There are three traditional architectures; DAS, SAN, and NAS. Direct Attached Storage (DAS) with its purchasing simplicity still dominates but as the volume of data and the number of users grow, Network Attached storage (NAS) and Storage area Networks (SAN) continue to make inroads into the SMB IT space. The improved storage, operational economics of SAN and NAS is what is stimulating the increased market share of these options with NAS and its simpler user experience evolving as the more popular. Unfortunately the realities of unpredictable but dynamic growth are beginning to compromise their benefits. However there is another emerging technology that is establishing credentials in the commercial space, clustered storage architectures, but that is the subject of another blog.

The object of this blog is to briefly comment on the traditional storage architecture, one normally referred to as monolithic and is an architecture that is independent of whether the storage is being implemented as a SAN or NAS. A monolithic architecture or in today’s topical parlance, scale-up, is an architecture that is characterized by storage that sits behind dedicated controllers or server heads. It is an architecture that dates back into the 1950’s when direct attached, on-line storage was in the nascent stages and over the years evolved to be the dominant storage architecture in today’s data center. This architecture has been very successful as the product lines of IBM, EMC, HDS, HP and the tier two folks such as 3PAR, Nexsan, Compellent, DDN and BlueArc will testify.

However despite its many attributes of a monolithic architecture it has five distinct disadvantages that creates significant angst for the SMB;

1. Tends to create inefficient “Islands of Storage” that increase cost and management complexity.
2. Creates bottlenecks that can challenge data access.
3. Storage is generally poorly utilized driving up the cost of usable storage.
4. Tends to be complex and have limited flexibility and disruptive scalability that drives defensive “poor” purchasing behavior.
5. Have inherent single points of failure that challenge reliability and data access.

These failings tend to drive complexity and cost which encourages poor purchasing behavior. It is not uncommon to witness storage administrator’s purchase more storage than is immediately required to avoid the significant hassle and risk associated with growth. Advanced purchasing while appreciated by vendors is the wrong way to purchase storage. If risk can be eliminated why pay today’s prices for tomorrow’s requirement incurring not only increasing capital cost but additional operational cost such as energy and cooling. While these are all characteristics that work against the storage consumer by introducing cost, complexity, and as noted, angst into the routine tasks of managing storage and surviving data growth there is a technology that promises a more effective approach.

Clustered storage appears to offer an interesting set of alternative attributes that directly address the failings of monolithic options and as noted earlier will be the subject of a future post.

Seven Attributes that Define Effective SMB Storage.

Read More >>

The storage challenges faced by small and medium sized companies are no different than those experienced by larger enterprises. While IT consumers in SMBs and larger enterprises have similar service levels expectations the SMB rarely has the specialized storage resources that larger businesses enjoy. SMB IT staff tend to be generalists with a broad portfolio of IT issues to manage and are unlikely to have the in-depth knowledge of any one specialization -- such as storage. Faced with the realities of limited staff, expertise and finances, an SMB is challenged with finding a storage solution that is cost effective, easy to implement, manage, maintain and one that will do the job So what is the SMB looking for in a storage solution? What are the attributes that would characterize an ideal solution for the SMB?

1. Simplicity: The concern that regularly percolates to the top is ease of install, and ease of normal daily use. If the complexity exceeds the capabilities of the resources available then failure is unavoidable. This suggests that the elimination of operational complexity is a must.
2. Affordability: Budget dollars are scarce so an affordable acquisition cost and low operational costs are critical. Flexibility and transparent upgradability directly impact operational and capital expenditures. If purchases can be delayed until needed, hence avoiding the costly practice of buying forward then significant savings can be realized.
3. Scalability: Any solution must be readily scalability in capacity and performance to accommodate data growth. It must also be scalable in time meaning that an existing investment can be “transparently” upgraded with new technology as it becomes available.
4. Flexibility: Flexibility protects the initial investment and is a key characteristic of a “buy as needed” solution.
5. Reliability: The burden of maintainability should be on the shoulders of the “wan-a-be” SMB solution vendors. Self maintaining and self healing technologies are a key characteristic of the ideal SMB storage solution and are a must for any meaningful SMB solution.
6. Data Integrity: While the value and criticality of business data is independent of enterprise size the severity of the consequences of data loss or corruption could be significantly different. Data protection technologies and methodologies that protect against data loss or compromised data integrity are a must for large and small enterprises but again operational simplicity for the SMB is key.
7. Appropriate Performance: While SMB’s may not have the aggressive performance demands common in large enterprise tier 1 transactional environments, they do have data access and data delivery requirements (IOPS and bandwidth) that have to match their business needs. In other words what they are looking for is appropriate performance, however that may be measured.

VMWorld delivers a pragmatic illustration of virtualizations powerful impact on green data center strategies.

Read More >>

Three fully functional virtualized data centers were assembled to support all the booth, keynote and lab activities at VMWorld with EMC, NetApp, HP, IBM, Cisco, Chip PC, MDS, and XSIGO contributing equipment and support.

To highlight the significance of virtualization as a primary component of a green data center strategy the VMware team presenting an interesting comparative analysis that looked pragmatically at the logistical differences between physical vs virtualized environments.

VMWorld was supported by a total of 37,248 VM hosted on 776 physical servers. There was 37TB of total memory and 348TB of total data center storage. The significance of the data center power consumption of 540 kWhr is appreciated when it is understood that if 37,248 physical servers were actually used the total power consumption would dramatically increase to approximately 25MWhr’s. Using these power consumption numbers and a $0.10/kWhr energy cost, virtualization dropped the hourly utility bill for the VMWorld data centers from $2500 to $54!

Reduced power costs were not the only operational savings. It was calculated that 2483 equipment racks would be needed to support a 37,248 server implementation but when virtualized this number shrinks to 28, radically dropping the floor space requirement from 154, 000 sq ft to 1700 sq ft.

Admittedly this was a bit of an artificial environment but the claim was that they were fully functional data centers and the numbers presented by the VMware team, accepted at face value, were impressive. The objective of this post is not to argue the detail of the numerical specifics but highlight the potential of virtualization as an enabler of greening the data center. Virtualization does not come for free and has its own unique set of issues and complexities, not the least being the management of what is characterized as VM sprawl. However, the numbers presented serve very well to illustrate the power of virtualization as a cornerstone of a data center efficiency strategy.

Perhaps this is why VMware currently counts 970 of the Fortune 1000 as customers.

Dell addresses the growing problem of electronic e-waste

Read More >>

In a recent press release Dell described their success in attacking the problem of e-waste. For those not conscious of the issue of e-waste, hazardous substances from discarded electronics that are populating landfills has become a significant issue with the EPA estimating that in 2005 there was 2.6M tons of e-waste made up in part from 133,000 personal computers being discarded each day in the US and 2.6 Billion units forecast to be scrapped this decade.

Dell has decided to address the issue directly with targeted programs with Staples and Goodwill. Staples has launched the program in 1500 stores and Goodwill are creating 1400 recycle drop-off points nationwide.

The results to date are to be commended with a 2008 achievement of 275M lbs of used computer equipment recycled and kept from contaminating the countries landfills.

The equipment is not the only concern. Dell reduced their packaging materials by 9.5 m LBS which is equivalent to approximately 150,000 trees. Innovative twist, they eliminated packaging material by using 2,000,000 recycled milk jugs (HDPE) as protective cushioning.

Dell has certainly displayed leadership with their corporate responsibility programs and it could be argued that they have the resources to afford such efforts. However these programs are not luxuries and all manufacturers large and small should be paying attention to sustainability practices. The problem has the attention of legislators with a state level movement gaining traction causing recycling fees to be imposed on manufactures at the point of sale. My understanding is that to date 19 states plus New York City have passed legislation mandating statewide e-waste recycling. Several more considering such in 2009 and 2010. All laws to date other than California use the producer responsibility approach where manufacturers must pay for recycling. California however passed a consumer, fee based, law.

So my message to vendors is to be prepared as in the words of Paul Revere “the regulators are coming”, my apologies to the historical purists. It is better to plan and be prepared than to ignore and scramble. Note to the end user; challenge your vendors on their end of life disposal plans if you do not you may be stuck with the cost of dealing with responsible recycling.