Per this blog’s opening “sound-byte”, if EMC’s ScaleIO virtual server SAN for block data were a car, it would most definitely be a Formula One (F1) Grand Prix racing car. I’m talking about being fast here…lightning fast in fact -- with seemingly unnatural speed, power, kinetic energy and agility. Hey, nothing goes around a closed circuit race track faster than a F1 racer, right?
[Apologies to you NASCAR fans out there, but to me a Grand Prix race course with its intricate and varied straight-a-ways; left and right curves and occasional tunnels make it most challenging to F1 drivers…aka, the “users”.]
I mean, I can’t believe how fast those drivers shift a hundred times or more per lap, delay braking, change direction instantly it seems (defying Gs) and accelerate hard out of curves, etc. at speeds in excess of 200MPH. Reminds me of my old “slot car” circuit days – 1/24 electrified plastic bodied cars locked in that middle track/lane groove...more or less sticking to the track. But I’m ‘fish tailing’ here…
So now that I’ve made this rather bold, somewhat self-serving analogy and set the stage (and tone) for some unabashed shout-outs for ScaleIO, it’s time to pay it off.
There’s a number of converged and hyper-converged software defined storage solutions out there from various vendors – including EMC. Some are sold in the form of bundled server SAN (Storage Area Network) hardware appliances. Others are sold as “pure” software with appropriate licensing. All have this in common, though: they use host server HW resources (and only host server resources) configured with local, direct attached storage (DAS) to form a virtual, aggregated pool of storage capacity. Hence, industry expert Wikibon coined the “Server SAN” term back in January 2014….just to describe this new emerging technology.
The beauty of a pure SDS system or HW server SAN appliance is that you leverage existing host servers’ CPU (i.e., ‘compute’), onboard connectivity (i.e., Ethernet) and storage (raw DAS capacity) resources. For instance, say you have 10 host servers in your data center or equipment room. Typically, those 10 servers are acting independently, separate from one another. Any data on each of those server’s flash or hard disk drives (usually at least one, most often two or more per server) is resident/written only on that server if it’s not backed up on an external storage – i.e., a SAN array connected by iSCSI or FC. So if ‘server A’ goes down/off-line for any reason, data on server A’s DAS device(s) is unavailable.
Happily, that “sad and lonely” data storage situation is eliminated by simply installing -- for example -- ScaleIO server SAN software modules on each of those participating host application servers with local DAS devices. Now you have not only a server cluster, but a virtual server SAN consolidated pool of connected servers that 1) linearly combines each server’s CPU I/O handling into the ‘power of one’; 2) aggregates each server’s raw excess DAS capacity into the ‘sum of all’; and 3) benefits from the resulting total I/O (IOPS) linearly scaled performance…and I/O bandwidth.
Equally important, those once isolated (and lonely) data blocks on each of those previously “isolated” host servers is now backed up across the new cluster pool. Distributed and dispersed data copies -- accessible and shared by all host nodes. That’s essentially the full measure and value of today’s contemporary converged, software defined server SAN – and how you can leverage (and monetize) previously unused/excess raw storage capacity without needing to install a specialized external storage array/resources.
But I digress again. Many of you already know what a SDS “server SAN” is and what it can do. But how can you tell if it’s ‘good’ or not? Back to my F1 race car highly stretched analogy. Below attempts to lay out -- albeit, with a somewhat tongue and cheek comparison of key attributes – my thinking.
Which do you think is faster?
Nodes = Cylinders
Individual ‘host server’ nodes forming a ScaleIO cluster can be thought of as the number of cylinders in the engine. Most will agree, the more pistons/cylinders you have working together in an engine, the more power and torque you get.
IOPS = RPMs
IOPS for the total cluster/pool is analogous to engine speed in my mind. More revs beget more performance from a given engine ‘configuration’…
MB/sec = MPH
Throughput/bandwidth can be thought of like actual speed (MB/sec = MPH)…the faster you go the more ground you cover in a given time slice.
Low Latency = Fast Acceleration
Latency is simple…its engine/throttle response and acceleration…as in ‘0 to 100’ in 4 seconds or less…when I key issue an I/O ‘throttle command’.
Raw Capacity = Fuel Tank Capacity
I think of raw capacity to being like how many gallons the car’s fuel tank can hold. Small tank? Better hope for a small track and short race/number of laps. Here, higher capacity means bigger endurance w/o needing ‘a pit stop’ to get extra fuel.
Elasticity = Number of Gear/Ratios
In this case, elasticity refers to ‘flexibility’ or ‘adaptability (to changing track and crace conditions) so I equate that to the number of available gears and/or type of tranny (as in, a F1 paddle shifter…) and the ability to add/remove resources, ‘throttle IOPS, move around data based on storage tiers/pools, data temperature, etc.
Agility = Suspension/Handling
Per above, F1 racing cars’ phenomenally crisp turning and track gripping handling are legendary….and the ability to be “agile on the fly” in response to changing user and application demands is a hallmark of SDS platforms – especially ScaleIO. Indeed, Elasticity and Agility are very close terms…but they each have subtle differences…and usually go hand-in-hand….to give you flexibility while negotiating dynamic track conditions.
Ease-of-management = Drivability
All the cylinders, raw performance, power, speed, available gears/rations and suspension/handling won’t do you an ounce of good if the car’s hard to drive, control and monitor.
Left: F1 Racer Steering Wheel/UI Right: ScaleIO Steering Wheel/GUI
Having fun yet? I’m almost done. Now for my pay-off on why ScaleIO is not only like a F1 Racer….but why it’s the very BEST SDS server SAN out there in the field.
As a F1 racer, would simply sweep the track…and brush off the competition. Why? No other SDS or hyper-converged server SAN (HW or SW) singularly can exceed – or even approach – ScaleIO in terms of:
1) #of Cylinders -- ScaleIO scales from a minimum of 3 to thousands of nodes.
2) Max RPMs -- ScaleIO linearly performance scales to many millions of IOPS.
3) Top MPH – Like max IOPS, ScaleIO’s MB/sec throughput capability leads the grid and posts the highest ‘MPH’ numbers.
4) Acceleration -- with RAM Caching; PCIe I/O Flash Card turbo charging and SSD device super-charging for high performance storage pools with ‘hot data’, ScaleIO’s low latency is in a class all to itself.
5) Fuel Capacity – if F1 cars were also measured by max gallons hey could carry …ScaleIO with its 16PB total raw capacity would handle the longest courses and most laps without stopping to refuel in the pits or add another ‘team car’.
6) Number of Gears – with its tiered storage pools (for white hot, hot, warm and cold data storage) – along with Protection Domains and Server Domain Fault Sets – ScaleIO has inherent superior Elasticity and many ‘gears’ to select.
7) Suspension/Handling -- ScaleIO is the Grand Prix Champion with its ability not only to add/remove server and DAS resources ‘on the fly’ without disrupting operation or reconfiguring cluster resources – but to also reconfigure cluster resources/operation without requiring fork lift upgrades, planning/provisioning or data migrations, etc.
8) Driveability – ScaleIO has a very simple, intuitively laid out GUI, “dashboard” and command interface – so it’s easy to start up, drive and manage on the track….by only one admin/FTE.
In closing, ScaleIO is hot and very fast. Grand Prix F1 Racer hot. And it’s highly scalable too. It has performance and capacity scalability second to none and blows the doors off its competitors. Moreover, it’s highly elastic, flexible and agile…for real world, ‘Daily Yeoman Street’ driving for your IT Datacenter. After all, what’s all this performance worth without proven reliability, quality of service and continuous/high availability (HA)? (Unless you’re a Formula One driver on the track). In that regard, ScaleIO combines the reliability of your high performance family sedan (or your trusty SUV/CUV) with that high strung F1 Racer thoroughbred I’m obviously impressed with.
But don’t take my word for it! Stay tuned for a follow-on blog in the very near future that presents actual lab test results on ScaleIO performed by the Storage Review experts. I think you’ll find their results “quite impressive and compelling”. Then you’ll get why I’m bragging on ScaleIO…and will also think of it as Formula One Race Car. One that you’ll want to take it around a few laps to see for yourself.
Turbo Charge Your Storage with ScaleIO Performance!!
ScaleIO Competitive Team