So Phil over at http://joeconsultant.blogspot.com/ has been chiding me that this blog is more about short snippets than real content. I disagree, but lately he is right in the fact that I have been real busy. My early posts about "The Spark" is mainly about the lack of true Fire in the industry.
Technology that is new ... First off I'd say that NAND Flash may not be "NEW". But it's use as a SSD (Solid State Disk) is a newer development. The throughput on this technology is pretty good. The IOPS you can get are about 10000 random read and about 600 write (SPECS). SLC has a much faster write but reads are still really good. I have a MLC SSD on my laptop it's not bad at all. Once the inital disk buffer gets full you can really tell where MLC fails vs SLC technology. I don't have an SLC drive to test against like a drive from http://www.rocketdisk.com/ the Mtron's are fast and expensive but the throughput is excellent. The Intel X25-E's and the new Samsung 256GB SSD are putting up good numbers as well. **In case anyone from Intel or Samsung or Rocketdisk are reading. Toss me an SLC to test against and I'll get it done :)**
So what is so important about this and why is the a game changer? Up until now to get the kind of IOPS 4 of these drives in RAID 10 can provide you would be looking at about 2 or so trays of 15K FC-AL disks in an EMC array. No worries about rotational latency anymore with NAND Flash. That's not bad but of course these small 2.5" SATA drives are not enterprise NOR dual connected (Like the FC-AL controllers). EMC has announced that NAND Flash drives are coming to the new CX4 Line and soon the NS120 and NS480. My EMC people tell me that NAND Flash drives have to be bought by the tray, that's 15 at a time. Cost? A cool 150K roughly! Goodness. This is a game changer due to the fact that now you don't have to buy a huge SAN array in order to get smoking performance for Enterprise SQL/Oracle installations anymore. Also the type of performance on these drives allows enterprise virtualization consultants to run some pretty nice VM teams in Workstation to simulate some small-scale environments to demo to customers. No more are you sitting there with your laptop and watching that 7200 RPM SATA drive churn to run both windows and VM's in order to demo to a client and win the sale. With many of the newer laptops you can run 2 2.5" SATA drives in them. The capacity on these types of drives isn't up to the 500GB yet ... but it's getting there. Samsung is going to release it's 256GB soon, so we are close enough to consider 2 of these in a laptop running RAID 1. Since Most VM's have a 75/25 read/write ratio the IOPS are there to run the VM's that you need to and your host. Having the responsiveness and the simulation ready might just be the thing to get that sale.
Phil and Tom's little discussion on Maturing IT is interesting. Phil points out HERE that such HOT technologies as virtualization have been around for years. I agree for servers and Operating Systems, but the next game changer that I am going to speak about is the virtualization of the network switching layer in the Data Center network.
Imagine that you have a vCloud level virtualization practice and you need to have the following routing protocols in a virtual environment : EIGRP, OSFP, BGPv4 and other advanced switching and routing concepts including unified I/O and FCoE. Is this truly new? It's debatable on some of the stuff. FCoE is indeed newer concetp-wise. If you leverage the Nexus1000V which is going to be built into the ESX hypervisor, the possibilities are almost endless. the complexities of the internal infrastructure become limited ONLY by your imagination. You could have multiple complete routed infrastructures that have complete separation and stateful fail-over. It's coming. Microsoft has been touting that 97% of the server market is still physical. If that is truly the case properly implementing VMware ESX 4.0 and Nexus1000V will be of MAJOR importance. So if you are a virtualization consultant and you only know OS, VMware, and Storage you might be in need of up-skilling. True high-end virtualization consultants are going to need to understand more and more as this accelerates forward. So spinning forward into 2009, we can expect to see all these services tied into the Virtual DataCenter OS. Augment these base DataCenter OS services with the new VMware View, ThinApp, and the LifeCycle control set and your DataCenter becomes easily manageable and controlled. You have everything covered and just need to worry about your supply pools : Storage, CPU, Memory, Network, and how to manage your growth. At last a DataCenter that you can live with and meet your SLA's with less effort.
You can then:
-Retire the employees that don't meet your needs be that Vision and/or skillset.
-Focus on the TRUE nature of IT, the elimination of itself.
-Proving ROI by greening the DataCeter and focusing on proven value.
-Prove value by responding to Client requests both Internal and External faster than ever before.
- All this and MORE.
Next time I promise I'll post more, PHIL :P.
No comments:
Post a Comment