A different angle on Datacenter Disruption

posted Nov 27, 2015, 7:09 AM by Niels Jensen   [ updated Nov 27, 2015, 8:21 AM ]
Yesterday Safepark attended Computerworld's How To "Server/storage/visualization" at Wihlborg's Conference Center in Ballerup on the outskirts of Copenhagen towards the Northwest. This day could best be characterized, at Datacenter Disruption from a different angle. The opening talk was given by storage architect lead Beat Balbier from Oracle in Switzerland. The message of his talk was, that Oracle hardware engineers and software engineers works side by side to co-engineer the create increase performance in the solutions Oracle offers its customers. He mentioned, that increased the capacity and performance of their storage system more than 50X by using Hyper Columnar Compression (HCC). He also said, that usually the cost of storage is one third OPEX and two third CAPEX. At the end of the talk he mentioned Oracle's (Almost) Zero Data Loss Recovery Appliance. To us it appears, as if Oracle continues to benefit from the acquisition of Sun Microsystems some yeas ago.

The second talk by engineer Martin Plesner-Jacobsen from Veeam changed the focus from storage to the business benefits of it, by focusing on availability. Their technology which are based on the hypervisor under the virtual server allows them to take a snapshot backup of a server, and then start that server from the snapshot, and hence proving, that the backup can be used in a recovery situation. Currently their solutions support WMware and Hyper-V, but they have announced support also for linux. Veeam already provide a free backup solution for your Windows desktops and laptops, which you can read more about and download here.

The third talk of the morning was Tom Christensen from Hitachi. His main message was that Hitachi has been working on flash for some years, and they will have storage class flash available before the end of the decade. However, he also talked about their Unified Compute Platform (UCP), which is available in four different sizes from UCP 1000 to UCP 6000. However, the pictures of these different UCP's under the text UCP X000 were identical except for some scaling. I wonder if the size of the box is the only difference between the UCP offerings from Hitachi?

Following a coffee and networking break manager Kim Sneftrup from Proact started with the old story of IT and the business. His talked focused on standardization and automation of the IT processes by creating a portal towards the business from which the IT services could be brokered. This talk left me wondering: Who decides when a new application is needed by the business? What about automating the business processes rather than just IT processes? There are both open source tools, such as Bonita BPM from Bonitasoft, and commercial offerings, such as Blueworks Live from IBM. You may also want to take a look at the draft redbook "Process Discovery Best Practices Using IBM Blueworks Live", which is available here.

The morning session were rounded off by two flash storage appliance providers: Tintri and Nimble Storage. The message from engineer Ultich Slothuus from Tintri was clear and simple "newer again manage storage". Their largest installation worldwide is currently 1.8 PB at Autodesk. The system involve 8 datacentres and 25.000 virtual servers, and this storage solution requirer 1 hour per week of maintenance work. Just too bad their solution don't scale down to my current home use needs of just 4-5 TB storage. Engineer Steve D'Amore from Nimble Storage had taken along a very satisfied customer to his presentation. The customer was Forca, which is a small Danish company administering pension payments. Their message was you just plug-in the Nimble Storage box and it just works. Forca was rather dissatisfied with their former supplier of storage. The message from Nimble Storage was that their solution reduce data center rack space by 90% with equivalent savings on cooling and electricity. Their smallest unit, the CS201, is almost fit for home usage.

After a networking lunch the afternoon kicked-off with engineer Jens Melhede from Violin Memory explaining why you should choose them as flash storage provider. Their solutions scales from 5 TB to 70 TB per 3U box with a 99.999% uptime for a 2 PB stack. Management interface to the system is created in HTML5. Violin Memory is partly owned by Toshiba, who has the rights to many patents in area of flash storage. The only flash storage vendor not represented at the conference appears to be Pure Storage.

The idea behind Nutanix is simple: Bring the performance and simplicity of management that drives the huge data-centers of Google, Facebook and others - so-called web-scale engineering - to smaller companies as turnkey systems. Nutanix runs on either WMware, Hyper-V or Acropolis (a KVM fork). The idea is to eliminate service windows, i.e. reduce downtime to the level of Google Apps and Facebook. I you are building a new data center or consolidating old ones, the you should at least have a look at Nutanix. The Nutanix extreme compute platforms are essentially a data center in a box - so there are 3 data centers on the left. On the outside the 3 data centers may look the same, but the inside could be quite different.

Dell also sent an architect to talk about flash, and the fact, that based on capacity shiped Dell is a very large storage vendor indeed. At the end of talk there were questions about the impact of the recently announced merger of Dell and Emc. The merger is expected to be complete sometime in 2016.

The final presentation of the conference was from RanTek, a Danish company located in Randers in middle part of Jutland. RanTek is a 15 year old company with a focus on IT optimization and performance. They have partnered with Riverbed to provide Danish customers access to that company's branch office solutions using the Steelfusion Zero Branch IT solution. One of their customers Grundfos rounded off the day by outlining their journey from 92 data centers to just three in Denmark (Bjerringbro), Singapore and USA. The hardware consolidation is processing well and will be completed by mid next year a head of schedule. Application consolidation have turned out to be much more demanding, and they haven't started looking at the production platform IT. People attempting to do something similar should be aware of the emotional moment when IT is removed from a location, and the personnel issues with such a global project - especially the efforts going into employee motivation - cost cutting is not very motivating. Grundfos IT in numbers: 5.5 PB structured data, 0.6 PB unstructured data, 6500 managed mobile phones, 15000 PCs, 18500 network ports, 2750 servers - +95% virtualized and 600+ employees.

All in all a good conference with some interesting appliance vendors, and a spectacular user story to round things off and touching on some of the human sides of a disrupting consolidation project.